Internships 2025-2026

The exponential advances in Machine Learning (ML) are leading to the deployment of Machine Learning models in constrained and embedded devices, to solve complex inference tasks. At the moment, to serve these tasks, there exist two main solutions: run the model on the end device, or send the request to a remote server. However, these solutions may not suit all the possible scenarios in terms of accuracy or inference time, requiring alternative solutions.

Cascade inference is an important technique for performing real-time and accurate inference given limited computing resources such as MEC servers. It combines more than two models to perform inference: a highly-accurate but expensive model with a low-accuracy but fast model, and determines whether the expensive model should make a prediction or not based on the confidence score of the fast model. A large pool of works exploited this solution. The first ones to propose a sequential combination of models were [1] for face detection tasks, then, in the context of deep learning, cascades have been applied in numerous tasks [2,3].  

Early Exit Networks take advantage of the fact that not all input samples are equally difficult to process, and thus invest a variable amount of computation based on the difficulty of the input and the prediction confidence of the Deep Neural Network [5]. Specifically, early-exit networks consist of a backbone architecture with additional exit heads (or classifiers) along its depth. At inference time, when a sample propagates through the through the network, it passes through the backbone and each of the exits in and the result that satisfies a predetermined criterion (exit policy) is (exit policy) is returned as the prediction output, bypassing the rest of the the rest of the model. In fact, the exit policy can also reflect the capabilities and load of the target device, and dynamically adapt the network to meet specific runtime requirements [6]. 

Our project is to use cascade models and/or early-exit models in the context of Edge Computing to improve the delay and reduce the resource usage of ML inference tasks at the edge. Of crucial importance for cascade models or early-exit models, is the confidence of the fast model. Indeed, if the prediction of the first model is used but wrong, it may lead to a low accuracy of the cascade model, even if the accuracy of the best model is very high. Similarly, if the first model confidence is set too low, it will never be used, and the computations will be higher than using only the second model by itself, additionally, we will use unnecessary network resources and have higher deals than necessary. Researchers have proposed methods to calibrate such systems [4]. However, they have not explored the choice of the loss function of such systems in depth.

In this project, we will explore the use of a new loss function for the fast models (or first exit) of cascade networks (of early-exit models). Indeed, such networks do not have the same goal as the global system, as they should only act as a first filter.

Useful Information:

The internship can be followed by a PhD for interested students. A PhD grant is already funded on the topic.  

Bibliography:  

[1] Viola, P., & Jones, M. (2001, December). Rapid object detection using a boosted cascade of simple features. In Proceedings of the 2001 IEEE computer society conference on computer vision and pattern recognition. CVPR 2001 (Vol. 1, pp. I-I). Ieee.

[2] Wang, X., Kondratyuk, D., Christiansen, E., Kitani, K. M., Alon, Y., & Eban, E. (2020). Wisdom of committees: An overlooked approach to faster and more accurate models. arXiv preprint arXiv:2012.01988. 

[3] Wang, X., Luo, Y., Crankshaw, D., Tumanov, A., Yu, F., & Gonzalez, J. E. (2017). Idk cascades: Fast deep learning by learning not to overthink. arXiv preprint arXiv:1706.00885.

[4] Enomoro, S., & Eda, T. (2021, May). Learning to cascade: Confidence calibration for improving the accuracy and computational cost of cascade inference systems. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 35, No. 8, pp. 7331-7339).

[5] Laskaridis, S., Kouris, A., & Lane, N. D. (2021, June). Adaptive inference through early-exit networks: Design, challenges and directions. In Proceedings of the 5th International Workshop on Embedded and Mobile Deep Learning (pp. 1-6). 

[6] Laskaridis, S., Venieris, S. I., Almeida, M., Leontiadis, I., & Lane, N. D. (2020, September). SPINN: synergistic progressive inference of neural networks over device and cloud. In Proceedings of the 26th annual international conference on mobile computing and networking (pp. 1-15).

2- Limiting Cloud and Network resource consumption and energy usage using Machine Learning (ML) Model compression.

Name: Frédéric Giroire et Davide Ferré

Mail: frederic.giroire@inria.fr

Web page: https://www-sop.inria.fr/members/Frederic.Giroire/

Place of the project:

Address: Inria, 2004 route de Lucioles, SOPHIA ANTIPOLIS

Team: COATI (common project Inria/I3S)

Web page: https://team.inria.fr/coati/

Pre-requisites: Knowledge in networking and machine learning. Python.

Description:  

In recent years, there has been a growing usage of Machine Learning (ML) models in cloud computing, contributing to the adoption of Machine Learning as a Service (MLaaS) [1], studied in several applications such as image recognition and self-driving cars [2]. Cloud and network operators have faced challenges in developing efficient strategies for utilizing computational resources to support machine learning tasks. Among these challenges, scheduling is an important one. A scheduler must determine on which machine each task is executed and its processing order. This becomes especially critical when tasks must adhere to deadline constraints.

In the context of saving computational resources, researchers have investigated neural network compression techniques, including pruning and quantization . However, these approaches typically involve compressing the model during the training phase, necessitating re-training the model after compression. Recent approaches compress neural networks at inference time [3, 28], reducing network size to varying degrees. Greater compression yields lower latency (i.e., the processing time of a task) but at the expense of accuracy.

In [3], we introduced a scheduling system using compressible neural networks for image classification tasks, in which several heterogeneous machines could be used. We developed an approximation algorithm with proven guarantees for maximizing the average accuracy while respecting deadlines constraints. In [4], we proposed scheduling algorithms to maximize accuracy while adhering to an energy budget constraint. Indeed, cloud and network operators are compelled to mitigate their cloud carbon footprint, driving researchers and scientists to investigate novel methods for conducting ML inference with greater energy efficiency. The adoption of MLaaS and the expanding size of neural network models has resulted in increased energy consumption, particularly during the inference stage [5]. According to reports from NVIDIA [9], 80-90% of Artificial Intelligence (AI) costs stem from inference. Indeed, processing millions of requests, such as those encountered in social networks, can lead to a significant number of inferences in deep learning models, resulting in elevated energy consumption and a sizable carbon footprint [6].

In this project, we will study how to extend the proposed solution using ML learning model compression to an edge computing context [7]. In this context, a model can be either executed in the cloud or closer to the user in its cell phone or in an antenna. We will propose optimization models and algorithms to find efficient solutions to schedule ML tasks.

Useful Information:

The TER can be followed by an internship and by a PhD for interested students. A PhD grant is already funded on the topic.  

References

[1] MauroRibeiro,KatarinaGrolinger,andMiriamAMCapretz.2015.Mlaas:Machine learning as a service. In 2015 IEEE 14th international conference on machine learning and applications (ICMLA). IEEE, 896–902.

[2] Suman Raj, Harshil Gupta, and Yogesh Simmhan. 2023. Scheduling dnn inferencing on edge and cloud for personalized uav fleets. In IEEE/ACM 23rd International Symposium on Cluster, Cloud and Internet Computing (CCGrid).

[3] T. da Silva Barros, F. Giroire, R. Aparicio-Pardo, S. Perennes, and E. Natale, “Scheduling with Fully Compressible Tasks: Application to Deep Learning Inference with Neural Network Compression,” in CCGRID 2024 - 24th IEEE/ACM international Symposium on Cluster, Cloud and Internet Computing, (Philadelphia, United States), IEEE/ACM, May 2024.

https://ieeexplore.ieee.org/document/10701353

[4] T. da Silva Barros, D. Ferre, F. Giroire, R. Aparicio-Pardo, and S. Perennes, “Scheduling Machine Learning Compressible Inference Tasks with Limited Energy Budget,” in ACM Digital Library, vol. 32 of ICPP ’24: Proceedings of the 53rd International Conference on Parallel Processing, (Gotland, Sweden), pp. 961 – 970, ACM, Aug. 2024.

https://hal.science/hal-04676376/document

[5] J. McDonald, B. Li, N. Frey, D. Tiwari, V. Gadepally, and S. Samsi. 2022. Great power, great responsibility: Recommendations for reducing energy for training language models. arXiv preprint arXiv:2205.09646 (2022).

[6] Radosvet Desislavov, Fernando Martínez-Plumed, and José Hernández-Orallo. 2023. Trends in AI inference energy consumption: Beyond the performance-vs- parameter laws of deep learning. Sustainable Computing: Informatics and Systems 38 (2023), 100857.

[7] Varghese, B., Wang, N., Barbhuiya, S., Kilpatrick, P., & Nikolopoulos, D. S. (2016, November). Challenges and opportunities in edge computing. In 2016 IEEE international conference on smart cloud (SmartCloud) (pp. 20-26). IEEE.

[8] F. Giroire, N. Huin, A. Tomassilli, and S. Pérennes. Data center scheduling with network tasks. in IEEE Transactions on Networking (ToN), 2025.

https://ieeexplore.ieee.org/document/11048391

3- Adaptive Control of Connectionless Quantum Networks

Who?

Name: APARICIO PARDO Ramon

Mail: raparicio@i3s.unice.fr

Telephone: 04 89 15 44 01

Web page: http://www.i3s.unice.fr/~raparicio/

Where?

Place of the project: Laboratoire I3S - UMR7271

Address: 2000 Route des Lucioles, BP 121, 06903 Valbonne

Team: COMRED/SIGNET

Web page: https://signet.i3s.univ-cotedazur.fr/

What?

Pre-requisites if any:

Languages:

- Python language (absolutely)

- Deep Learning libraries (like TensorFlow [6], Keras) (recommended)

Theory:

- Machine Learning, Data Science, particularly Neural Networks theory (recommended)

- Classical optimisation theory (Linear Programming, Dual Optimisation, Gradient Optimisation, Combinatorial Optimization, Graph Theory) (recommended)

Technology:

- Computer networking (recommended)

- Quantum computing and networking (not necessary but convenient)

Detailed description: indicate the context of the work, what is

expected from the intern, what will be the outcome (software,

publication, …).

1. Objective and Impact

The objective is to develop a scalable, decentralized, and adaptive routing protocol capable of managing the flow of entanglement requests—termed Q-datagrams [1]—in large-scale, dynamic networks composed of terrestrial and satellite nodes. The core challenge is applying Reinforcement Learning (RL) to minimize the decoherence time of quantum states in memory queues while maximizing the entanglement throughput. This will be achieved by implementing a DQN-based Q-routing solution, leveraging recent success in applying this technique to classical communication networks environments [2-3].

The project will provide a robust, real-time adaptive routing framework essential for the scalable operation of the future Quantum Internet. By demonstrating the synergy between the decentralized nature of DQN-based Q-routing and the packet-switched Q-datagram paradigm [1], the solution will show improved efficiency and resilience against fluctuating network congestion and unpredictable link dynamics, outperforming static routing policies.

2. State of the Art (SOTA) and Context

The operational framework is grounded in modeling the Quantum Internet as a best-effort, packet-switched network, where entanglement requests are abstracted as Q-datagrams [1]. This abstraction is crucial as it allows established classical network protocols for routing, congestion and queue management to be applied to protect fragile quantum memory. The network environment is inherently dynamic and heterogeneous, encompassing terrestrial and satellite links subject to rapid changes, which renders static routing ineffective [4].

The routing solution is built upon the foundation of Reinforcement Learning (RL):

• Classical Q-Routing: The original Q-routing algorithm [5] is a decentralized technique where nodes learn optimal path costs locally. However, its state space representation scales poorly with large networks.

• Deep Q-Networks (DQN)-based Q-routing: To overcome the scaling issues and handle the complexity of quantum network states, DQN [6] uses neural networks to approximate the Q-function. Recent SOTA confirms that applying DQN to communication networks optimizes end-to-end performance [2-3]. This approach allows the routing decision to adapt autonomously to real-time network conditions (e.g., memory occupancy and link quality).

The primary focus of this research is to deeply integrate the DQN-based Q-routing approach with the Q-datagram model, ensuring that the decentralized learning process directly prioritizes paths that minimize entanglement swapping latency and decoherence.

3. Methodology and Work Plan (6 Months)

The research will be conducted in three distinct phases: simulation environment modeling, algorithm implementation, and comparative evaluation.

Phase 1: Simulation Environment Modeling (Months 1–2)

This phase involves establishing the simulation environment by building a discrete-event simulator for a Satellite Terrestrial Quantum Network topology (satellite and terrestrial nodes) where quantum memory queues and decoherence (fidelity loss as a function of time) are accurately modeled, formulating a cost function based on latency and successful entanglement delivery to guide the Deep Reinforcement Learning agent.

Phase 2: DQN-Based Q-Routing Implementation (Months 3–4)

This phase focus on the core algorithmic development by implementing the decentralized Q-routing engine at each node, replacing the traditional Q-table[5] with a Deep Q-Network (DQN) architecture [6] (as in [2-3]) to handle the complex state-action space, and training this network to learn the optimal routing policy based on the local observation of link congestion and the reward signal from successful Q-datagram deliveries.

Phase 3: Simulation and Evaluation (Months 5–6)

The final phase focuses on rigorous validation through conducting extensive simulations across various load and link failure scenarios, performing a comparative analysis of the DQN-based Q-routing performance against standard baselines like Shortest Path, and ultimately generating a comprehensive final report detailing the models, the performance metrics, and the final conclusions.

References: set of bibliographical references (article, books, white papers, etc) to be read by the student before starting to work on this subject

[1] L. Bacciottini et al., "Leveraging Internet Principles to Build a Quantum Network," in IEEE Network, doi: 10.1109/MNET.2025.3569494

[2] X. You, X. Li, Y. Xu, H. Feng, J. Zhao and H. Yan, "Toward Packet Routing With Fully Distributed Multiagent Deep Reinforcement Learning," in IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 52, no. 2, pp. 855-868, Feb. 2022, doi: 10.1109/TSMC.2020.3012832

[3] R. A. Alliche, R. A. Pardo and L. Sassatelli, "O-DQR: A Multi-Agent Deep Reinforcement Learning for Multihop Routing in Overlay Networks," in IEEE Transactions on Network and Service Management, vol. 22, no. 1, pp. 439-455, Feb. 2025, doi: 10.1109/TNSM.2024.3485196.

[4] Y. Zhang, Y. Gong, L. Fan, Y. Wang, Z. Han and Y. Guo, "Efficient Entanglement Routing for Satellite-Aerial-Terrestrial Quantum Networks," 2025 34th International Conference on Computer Communications and Networks (ICCCN), Tokyo, Japan, 2025, pp. 1-9, doi: 10.1109/ICCCN65249.2025.11133770.

[5] J. Boyan and M. Littman, “Packet routing in dynamically changing networks: A reinforcement learning approach,” in Proc. Adv. Neural Inf. Process. Syst., vol. 6, 1993, pp. 1–8.

[6] V. Mnih, “Human-level control through deep reinforcement learning,” Nature, vol. 518, no. 7540, pp. 529–533, 2015.

4- Fault-tolerant multiparty session types

Étienne Lozes and Cinzia Di Giusto

etienne.lozes@univ-cotedazur.fr

http://webusers.i3s.unice.fr/~elozes

I3S Lab, SCALE Team : https://scale.i3s.unice.fr/

Prerequisites: lectures on finite state automata

Keywords: distributed synthesis

Description in french with a picture: https://etiloz.github.io/visite-ens-lyon-sophia-2025.github.io/stages/12_Faulty_ST.html

Description in english:

Multiparty session types are finite automata used to globally specify a message-exchange protocol between multiple agents.

More precisely, a “global” type is a finite automaton whose transitions are labelled with arrows (p \to q : m) (agent (p) sends message (m) to agent (q)).

The set of arrows forms a partially commutative alphabet. An equivalence class of a “word of arrows” is called a Message Sequence Chart (MSC).

A global type is projected onto each agent (p) as a “local” type, which is a communicating automaton labelled with send actions (!m) and receive actions (?m).

The parallel composition of all these local types is the projected system associated with the global type.

The projected system also accepts Message Sequence Charts: each local system checks that the sequence of actions of its process belongs to its language.

A global type is said to be realisable if its projected system recognises only the Message Sequence Charts of the global type.

This is currently the fundamental criterion that a global type must satisfy to be considered “safe” (see for example [1, 2]).

This internship aims to explore the notion of global types beyond realisability.

A global type that is not realisable leads to a projected system that recognises Message Sequence Charts that fall outside the language of the global type,

corresponding to executions in which the agents have become “desynchronised”.

Rather than giving up on such a global type, we seek to add correction mechanisms to the projected system, either by adding information to the messages

sent in order to prevent desynchronisation (this is a known but complex construction, known as Zielonka automata [3, 4, 5]), or by “repairing’’ the

desynchronisation once detected, by adding “failure-recovery” messages enabling agents to agree either to “roll back’’ the protocol to a

point where they were not desynchronised, or to resynchronise with a process.

The internship may lead either to a bibliographical study on constructions of Zielonka automata, starting from their recent simplified

variants, or to an exploration of the second, entirely unexplored approach based on correction messages.

[1] Complete Multiparty Session Type Projection with Automata — Elaine Li, Felix Stutz, Thomas Wies & Damien Zufferey. CAV’23

[2] Realisability and Complementability of Multiparty Session Types — Cinzia Di Giusto, Étienne Lozes, Pascal Urso. PPDP’25

[3] Synthesising Asynchronous Automata from Fair Specifications — Béatrice Bérard, Benjamin Monmege, B. Srivathsan, Arnab Sur

[4] Notes on finite asynchronous automata — Wieslaw Zielonka. Theoretical Informatics and Applications, 21(2):99–135, 1987

[5] Optimal Zielonka-type construction of deterministic asynchronous automata — Blaise Genest, Hugo Gimbert, Anca Muscholl, Igor Walukiewicz. ICALP’10

5- Investigating Vulnerabilities and Opportunities of GPT‑based AI Assistants in the Peer‑Review Process

Name: Chuan Xu

Mail: chuan.xu@inria.fr

Web page: sites.google.com/view/chuanxu

Place of the project: Inria

Address: 2004, route des Lucioles, Valbonne, France

Team: COATI

Webpage: https://team.inria.fr/coati/

Pre-requisites : Experience of using machine learning programming framework such as PyTorch, Familiar with Hugging face

Description:
The machine‑learning community is increasingly adopting AI review assistants to cope with a growing volume of submissions and with reviews produced under time pressure. These systems can improve throughput and consistency, but they may also introduce new vulnerabilities (e.g., being misled by adversarial text) or hidden biases. This internship studies both risks and potential improvements introduced by GPT‑style review aids.

Objectives

1. Map and synthesize the current AI‑assisted peer‑review workflows and the relevant literature.

2. Empirically evaluate how GPT‑based reviewers respond to changes in a paper’s content and structure.

3. Assess vulnerabilities and possible misuse (e.g., hiding inconsistencies, “hacking” paper wording to obtain better reviews).

Research Questions :

· Which sections of a paper (abstract, introduction, results, conclusion, figures) most strongly influence GPT‑based review judgments?

· To what extent can inconsistencies or flawed arguments be concealed from GPT‑based reviewers while still appearing convincing?

· Can small, targeted changes in phrasing or structure increase the likelihood of a favorable AI review for unchanged scientific content?

References:

Overview of AI Review System in AAAI 2026 https://aaai.org/wp-content/uploads/2025/08/FAQ-for-the-AI-Assisted-Peer-Review-Process-Pilot-Program.pdf

6- Exploring memory performance for large Python computations

Who?

Name: Damien Saucez

Mail: damien.saucez@inria.fr

Web page: https://team.inria.fr/diana/team-members/damien-saucez/

Name: Arnaud Legout

Mail: arnaud.legout@inria.fr

Web page: https://www-sop.inria.fr/members/Arnaud.Legout/

Where?

Place of the project: DIANA team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: DIANA

Web page: https://team.inria.fr/diana/

Pre-requisites if any: Familiar with Python, Linux, basic system performance knowledge, highly

motivated to work in a research environment and exited to tackle hard problems.

Description:

Making datascience is not only a programming or machine learning issue, it is also

a system issue for most practical use cases. One of which is to make the best use of the available RAM.

Making computations requiring a large amount of memory that exceeds the available RAM

require the OS the swap memory pages on disk. Even in case your process does not exceed the

available RAM, the Linux memory management may proactively swap memory pages.

We observed that under certain circumstances, the swap process dramatically

reduces the performance leading to a pathological behavior in which retrieving pages

from the swap becomes much slower than the disk speed.

The goal of this internship is to understand and document the current Linux memory management,

how it interacts with the Python interpreter, and how to reproduce the

circumstances under which we enter a pathological behavior.

Excellent students will also have the opportunity to study the impact of virtualization

running multiple memory intensive workloads.

This internship requires a good understanding of the internals of the Linux operating system and

a good knowledge of C and Python. It will be mandatory to look at Linux

and Python source code (written in C) to understand the details and the undocumented

behavior.

Also a part of the internship will be to run experiments to reproduce and understand the conditions

under which we observe a pathological behavior.

Excellent students will have the possibility to continue for a Ph.D. thesis.

Useful Information/Bibliography:

What every programmer should know about memory

https://lwn.net/Articles/250967/

Python memory management

https://docs.python.org/3/c-api/memory.html

Memory Management

https://docs.kernel.org/admin-guide/mm/index.html

7- Impact of Large Language Model on the Cognitive Process

Who?

Name: Arnaud Legout

Mail: arnaud.legout@inria.fr

Web page: https://www-sop.inria.fr/members/Arnaud.Legout/

Where?

Place of the project: DIANA team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: DIANA

Web page: https://team.inria.fr/diana/

Pre-requisites if any: Python and web development, basic knowledge on how LLM work,

open to trans-disciplinary works, eagerness to work at the edge of computer science and

social psychology

Description:

Large Language Models (LLM) have been a revolution in AI in the past two years.

In particular, copilots (such as GitHub copilot) are used to assist humans by predicting

what they want to do next. However, little is known on how a copilot interact

with the cognitive process. In particular, is it possible for a copilot to influence

and even change the mind of the assisted human?

The goal of this Internship is to design, run, and analyze experiments on the

impact of a copilot on the cognitive process. The design task will be to refine

the research question we want answer and create an experiment that allows

to answer that question. For instance, assume you want to show that a copilot

can influence basic calculation tasks, we might design an experiment where

the copilot answers at random an incorrect result. The run tasks consists in

conducting the experiment with real persons that we will recruit among the

population of the students of the faculties involved in the project. The analyze

tasks consists in making a statistical evaluation of the results using resampling

techniques such as permutation tests and bootstrap confidence intervals.

During this internship, you will learn how to define a research question,

how run a research experiment, and how to make the statistical interpretation

of an experiment.

This internship might be followed for excellent and motivated students by a Ph.D. thesis.

Useful Information/Bibliography:

Erik Jones and Jacob Steinhardt. “Capturing failures of large language models via human cognitive biases”. In:

Advances in Neural Information Processing Systems 35 (2022), pp. 11785–11799.

Enkelejda Kasneci et al. “ChatGPT for good? On opportunities and challenges of large language models for

education”. In: Learning and individual differences 103 (2023), p. 102274.

Celeste Kidd and Abeba Birhane. “How AI can distort human beliefs”. In: Science 380.6651 (2023), pp. 1222–1223

Bill Thompson and Thomas L Griffiths. “Human biases limit cumulative innovation”. In: Proceedings of the Royal

Society B 288.1946 (2021), p. 20202752.

Canyu Chen and Kai Shu. “Can LLM-Generated Misinformation Be Detected?” In: arXiv preprint arXiv:2309.13788

(2023).

8- Enhancing Large-Scale Distributed Caching Systems with Erasure Coding

Supervisors: Sara Alouf and Kyrylo Tymchenko (Inria NEO team)

Mail: sara.alouf@inria.fr, kyrylo.tymchenko@inria.fr

Place: Neo Team, Inria, Sophia Antipolis

Pre-requisites: Knowledge in networking and probability theory, solid programming skills are a plus

Description:

In recent years, there has been a steady evolution of edge computing, storage, content delivery, and AI services, all moving closer to end-users [1]. This shift is primarily driven by the need to reduce latency, limit bandwidth consumption, and improve the overall throughput and responsiveness of the distributed systems.

In edge environments, caching plays a vital role in reducing latency and avoiding redundant requests to origin servers. However, due to intense performance demands, high request volumes, and inevitable hardware failures, server unavailability events are a common occurrence [2]. For example, traces from Akamai’s CDN indicate an average of 5.6 such events per cluster per day [3]. These incidents are typically short-lived but highly disruptive, triggering spikes in cache miss rates and increasing average latency [3]. The result is a degraded quality of service (QoS) and a diminished user experience.

A further consequence of transient server failures is load imbalance, often due to bucket assignment algorithms that co-locate requests by vendor or content group [4]. When one or more servers go offline, their load is redistributed unevenly, increasing eviction rates and accelerating SSD wear on the remaining nodes. This degrades both efficiency and maintenance costs.

To mitigate these problems, selective replication—based on content popularity and size—is widely used in industry [5]. Replication reduces cache miss spikes but comes at a high storage cost and can worsen load imbalance, limiting scalability and cost-effectiveness.

Erasure coding is a widely used space efficient alternative to replication [6, 7, 8]. Traditionally used in storage systems [9] and network communication for error correction [10], erasure coding has recently become more viable in caching and content delivery contexts due to advances in computational optimization [3, 11, 12]. Studies show that combining erasure coding with smart fragment placement and load-balancing algorithms can significantly improve fault tolerance, reducing cache miss spikes and I/O imbalance.

Yet, the application of erasure coding in edge environments is still relatively underexplored, leaving many optimization opportunities open. For example, placement algorithms that minimize both read and write imbalance could lower latency while reducing SSD wear. Similarly, adaptive coding schemes that adjust redundancy based on content popularity could balance performance and storage overhead dynamically, leaving more space for useful data.

The project will begin with a study of the current state of the art in the literature. The next step will be to explore the strategies to improve load balancing, storage overhead and fault tolerance. Finally, the student will develop a framework to test the performance of these strategies under different types of workloads.

References

Weisong Shi, Jie Cao, Quan Zhang, Youhuizi Li, and Lanyu Xu. Edge computing: Vision and challenges. IEEE internet of things journal, 3(5):637–646, 2016.
Erik Nygren, Ramesh K Sitaraman, and Jennifer Sun. The akamai network: a platform for high-performance internet applications. ACM SIGOPS Operating Systems Review, 44(3):2–19, 2010.
Juncheng Yang, Anirudh Sabnis, Daniel S Berger, KV Rashmi, and Ramesh K Sitaraman. C2DN: How to harness erasure codes at the edge for efficient content delivery. In 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI 22), pages 1159–1177, 2022.
Bruce M Maggs and Ramesh K Sitaraman. Algorithmic nuggets in content delivery. ACM SIGCOMM Computer Communication Review, 45(3):52–66, 2015.
Ganesh Ananthanarayanan, Sameer Agarwal, Srikanth Kandula, Albert Greenberg, Ion Stoica, Duke Harlan, and Ed Harris. Scarlett: coping with skewed content popularity in mapreduce clusters. In Proceedings of the sixth conference on Computer systems, pages 287–300, 2011.
Cheng Huang, Huseyin Simitci, Yikang Xu, Aaron Ogus, Brad Calder, Parikshit Gopalan, Jin Li, and Sergey Yekhanin. Erasure coding in windows azure storage. In 2012 USENIX Annual Technical Conference (USENIX ATC 12), pages 15–26, 2012.
Saurabh Kadekodi, Francisco Maturana, Suhas Jayaram Subramanya, Juncheng Yang, KV Rashmi, and Gregory R Ganger. PACEMAKER: Avoiding HeART attacks in storage clusters with disk-adaptive redundancy. In 14th USENIX Symposium on Operating Systems Design and Implementation (OSDI 20), pages 369–385, 2020.
Subramanian Muralidhar, Wyatt Lloyd, Sabyasachi Roy, Cory Hill, Ernest Lin, Weiwen Liu, Satadru Pan, Shiva Shankar, Viswanath Sivakumar, Linpeng Tang, et al. f4: Facebook’s warm BLOB storage system. In 11th USENIX Symposium on Operating Systems Design and Implementation (OSDI 14), pages 383–398, 2014.
Jad Darrous and Shadi Ibrahim. Understanding the performance of erasure codes in hadoop distributed file system. In Proceedings of the Workshop on Challenges and Opportunities of Efficient and Performant Storage Systems, pages 24–32, 2022.
Bernard Fong, Predrag B Rapajic, Guan Y Hong, and Alvis Cheuk M Fong. Forward error correction with reed-solomon codes for wearable computers. IEEE Transactions on Consumer Electronics, 49(4):917–921, 2003.
KV Rashmi, Mosharaf Chowdhury, Jack Kosaian, Ion Stoica, and Kannan Ramchandran. EC-Cache:Load-Balanced, Low-Latency cluster caching with online erasure coding. In 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI 16), pages 401-417, 2016.
Hai Jin, Ruikun Luo, Qiang He, Song Wu, Zilai Zeng, and Xiaoyu Xia. Cost-effective data placement in edge storage systems with erasure code. IEEE Transactions on Services Computing, 16(2):1039–1050, 2022.

9- End-to-End Latency Optimization in Open-Source 5G for URLLC

Who?

Name: Thierry Turletti and Walid Dabbous and Chadi Barakat

Mail: : Thierry.Turletti@inria.fr and Walid.Dabbous@inria.fr and Chadi.Barakat@inria.fr

Telephone: +33 4 92 38 78 79

Web page: https://team.inria.fr/diana/team-members/thierry-turletti/ and https://team.inria.fr/diana/team-members/walid-dabbous/ and https://team.inria.fr/diana/team-members/chadi/

Where?

Place of the project: Inria centre at Université Côte d'Azur

Address: 2004, route des lucioles, 06902 Sophia Antipolis, France

Team: Diana

Web page: https://team.inria.fr/diana/

What?

Pre-requisites if any: Solid background in networking and wireless technologies, strong skills in system programming (C/C++, python, shell).

Detailed description: indicate the context of the work, what is expected from the intern, what will be the outcome (software, publication, ...).

Ultra-Reliable Low-Latency Communications (URLLC) are essential for emerging 5G use cases such as industrial automation, autonomous vehicles, and remote medical applications. Achieving millisecond-level end-to-end latency requires a deep understanding of both 5G protocols and real-time software optimization.

This internship will focus on state-of-the-art open-source 5G stacks, specifically OAI5G [3] and srsRAN [4], to study, implement, and evaluate latency-reduction techniques. The first step will consist of a literature and state-of-the-art review, with particular attention to two recent studies [1,2]. After this first step, the student will then:

Analyze and reproduce the results reported in these studies.
Identify performance bottlenecks in both the OAI5G and srsRAN stacks.
Investigate how to combine the various optimizations proposed in the literature to maximize the performance of 5G stacks. Expected optimizations include those based on fine-tuning system parameters, as well as system-level packet-acceleration techniques such as eBPF, XDP, and AF_XDP.
Conduct experiments on a state-of-the-art 5G testbed, measuring real-time network performance and end-to-end latency improvements.

Expected outcomes include a comprehensive evaluation of latency-optimization techniques, a comparative study of OAI5G and srsRAN, and practical guidelines for applying these methods to achieve URLLC-grade performance in open-source 5G networks.

This internship is proposed within the context of the SLICES [5] European project and the national Priority Research Programme and Equipment (PEPR) on 5G, 6G, and Networks of the Future [6]. It is ideal for students passionate about wireless communications, real-time systems, and experimental 5G research, combining theoretical study with hands-on implementation and performance evaluation.

If secured funding becomes available (e.g., through research projects or scholarships), this internship may be extended into a PhD thesis for motivated students who demonstrate strong skills in conducting the planned experiments and analyzing the results.

References: set of bibliographical references (article, books, white papers, etc) to be read by the student before starting to work on this subject

[1] T. Tsourdinis, N. Makris, T. Korakis and S. Fdida, "Demystifying URLLC in Real-World 5G Networks: An End-to-End Experimental Evaluation," GLOBECOM 2024 - 2024 IEEE Global Communications Conference, Cape Town, South Africa, 2024, pp. 2954-2959, doi: 10.1109/GLOBECOM52923.2024.10901776.

[2] Gong, A., Maghsoudnia, A., Cannatà, R., Vlad, E., Lomba, N. L., Dumitriu, D. M., & Hassanieh, H. (2025, September). Towards URLLC with Open-Source 5G Software. In Proceedings of the 1st Workshop on Open Research Infrastructures and Toolkits for 6G (pp. 7-14). https://doi.org/10.1145/3750718.3750743

[3] OpenAirInterface, https://openairinterface.org/

[4] srsRAN project https://www.srslte.com/

[5] ESFRI SLICES European project, https://www.slices-ri.eu/what-is-esfri/

[6] PEPR on 5G, 6G and Networks of the Future, https://pepr-futurenetworks.fr/en/home/

10- Experimenting with Error Recovery Mechanisms in 5G Networks

Who?

Name: Chadi Barakat and Thierry Turletti

Mail: : Chadi.Barakat@inria.fr and Thierry.Turletti@inria.fr

Telephone: +33492387594

Web page: https://team.inria.fr/diana/team-members/chadi/ and https://team.inria.fr/diana/team-members/thierry-turletti/

Where?

Place of the project: Inria centre at Université Côte d'Azur

Address: 2004, route des lucioles, 06902 Sophia Antipolis, France

Team: Diana

Web page: https://team.inria.fr/diana/

What?

Pre-requisites if any: Solid background in networking and wireless technologies, strong skills in system programming (C/C++, python, shell).

Detailed description: indicate the context of the work, what is expected from the intern, what will be the outcome (software, publication, ...).

Cellular 5G networks deploy various mechanisms to combat wireless errors, ranging from Hybrid ARQ retransmissions to correct corrupted frames, to adaptive modulation to improve link quality and reduce the block error rate. These mechanisms are tuned based on real-time reports from UEs (User Equipment) about their network conditions (e.g., channel state information, block error rate, SINR) [1,2,3,4]. Given the diversity of wireless environments experienced by different UEs, and the various forms that noise can take (e.g., white versus correlated noise, or noise affecting only parts of the spectrum), it is crucial that UEs are allocated the best-quality subcarriers based on their individual conditions to maximize performance, independently of the conditions experienced by others. It is also essential that the allocation of frames to resource blocks be optimized to maximize total throughput across the wireless link, under any distribution of noise across UEs and subcarriers. Achieving this goal requires (i) precise feedback from the UEs about their wireless conditions, both in time and frequency, and (ii) a MAC scheduler that fully leverages this information in its resource-allocation decisions. We believe that a solid understanding has not been yet reached regarding the relationship between the precision of UE feedback and the scheduler’s ability to optimally allocate resources (considering varying device loads and the heterogeneous wireless conditions experienced by different UEs).

The goal of this internship is to delve deeper into this problem through extensive experimentation and 5G protocol analysis. First, we aim to study how error-recovery mechanisms are implemented in current open-source platforms such as OAI [5] and srsRAN [6], and to conduct experiments evaluating the effectiveness of these mechanisms in achieving their intended objectives. The work will involve running experiments on our SLICES-RI platform [7] and in our R2Lab anechoic chamber [8]. By introducing artificial wireless noise in different forms and running Internet traffic in both directions with multiple UEs, we will assess how efficiently the deployed mechanisms exploit the wireless link in terms of achieved throughput, and whether any bias in resource allocation may lead to fairness issues among UEs. We will also evaluate whether the deployed link-monitoring mechanisms provide both the base station and the UEs with an accurate view of the wireless channel and of the condition of the individual subcarriers that compose it, for example using the tool described in [9]. In the second phase, we will design experimental scenarios involving multiple UEs with different wireless conditions and traffic demands, and analyze how the gNB’s MAC scheduler allocates resources among them. In particular, we will assess the scheduler’s ability to leverage UE-reported channel information to allocate wireless resources in an optimal manner, taking into account both the wireless conditions experienced by each UE and their traffic requirements.

References: set of bibliographical references (article, books, white papers, etc) to be read by the student before starting to work on this subject

[1] Sesha Sai Rakesh Jonnavithula, Ish Kumar Jain, and Dinesh Bharadia. 2024. MIMO-RIC: RAN Intelligent Controller for MIMO xApps. In Proceedings of the 30th Annual International Conference on Mobile Computing and Networking (ACM MobiCom '24).

[2] Xenofon Foukas, Bozidar Radunovic, Matthew Balkwill, and Zhihua Lai. 2023. Taking 5G RAN Analytics and Control to a New Level. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking (ACM MobiCom '23).

[3] N. Saha, N. Shahriar, M. Sulaiman, N. Limam, R. Boutaba and A. Saleh, "Monarch: Monitoring Architecture for 5G and Beyond Network Slices," in IEEE Transactions on Network and Service Management, vol. 22, no. 1, pp. 777-790, Feb. 2025.

[4] A. Kak, V. -Q. Pham, H. -T. Thieu and N. Choi, "RANSight: Programmable Telemetry for Next-Generation Open Radio Access Networks," GLOBECOM 2023 - 2023 IEEE Global Communications Conference, Kuala Lumpur, Malaysia, 2023, pp. 5391-5396.

[5] OpenAirInterface, https://openairinterface.eurecom.fr/

[6] srsRAN, Project Open Source RAN, https://www.srslte.com/

[7] SLICES-RI, Scientific Large Scale Infrastructure for Computing/Communication Experimental Studie, https://www.slices-ri.eu/

[8] R2lab anechoic chamber, https://r2lab.inria.fr/

[9] Sesha Sai Rakesh Jonnavithula, Ish Kumar Jain, and Dinesh Bharadia. 2024. BeamArmor5G: Demonstrating MIMO Anti-Jamming and Localization with srsRAN 5G Stack. In Proceedings of the 30th Annual International Conference on Mobile Computing and Networking (ACM MobiCom '24).

11- Evaluating Perturbation Strategies Against Data Reconstruction Attacks in Federated Learning

Name: Francesco Diana, Chuan Xu

mail: francesco.diana@inria.fr, chuan.xu@inria.fr

Webpage: sites.google.com/view/chuanxu

Place of the project: Inria

Address: 2004, route des Lucioles, Valbonne, France

Team: COATI

Webpage: https://team.inria.fr/coati/

Pre-requisites : We are looking for a candidate with good analytical skills and coding experience in PyTorch.

Description:

Federated Learning (FL) [1] is a distributed paradigm designed to enable collaborative training of machine learning (ML) models while protecting user privacy by keeping sensitive training data locally on edge devices.

However, the assumption of strong privacy protection in FL is severely challenged by Gradient Inversion Attacks (GIAs) [2][3][4], also known as Data Reconstruction Attacks (DRAs)[5]. These attacks allow an adversary to recover sensitive training samples from the gradients or model updates exchanged by clients.

Two primary challenges arise when developing mitigation strategies:

1. Privacy Protection: ensuring that highly sensitive training samples are unrecoverable by the adversary.

2. Utility Preservation: guaranteeing that the defense mechanism does not significantly degrade model performance, increase training time, or compromise convergence.

Traditional methods, such as adding noise to gradients guided by Differential Privacy (DP) [6], have been widely explored [7]. Unfortunately, to achieve meaningful privacy guarantees, DP often requires adding a significant magnitude of noise, which severely impairs the final model accuracy, leading to an unsatisfactory trade-off between privacy and usability.

Hence, there is a pressing need for innovative defense mechanisms that actively protect data leakage without compromising model utility.

The goal of this project to explore and evaluate alternative defense mechanisms against DRAs that rely on modifications of data [8] or model representations/parameters [9][10][11] while preserving model accuracy.

For this internship, we expect the student to:

- Familiarize himself/herself with Federated Learning, attack threat models and Data Reconstruction Attacks.

- Evaluate the detectability of these DRAs.

- Implement and evaluate existing defense methods against different DRAs.

References:

[1] McMahan et al, Communication-Efficient Learning of Deep Networks from Decentralized Data, AISTATS 2017, pages 1273-1282

[2] Geiping et al, Inverting Gradients – How Easy is it to Break Privacy in Federated Learning?, NeurIPS 2020, pages 1421–1431

[3] Boenisch et al, When the Curious Abandon Honesty: Federated Learning Is Not Private, IEEE EuroS&P 2023, pages 175–199

[4] Carletti et al, SoK: Gradient Inversion Attacks in Federated Learning, USENIX Security 2025

[5] Diana et al, Cutting Through Privacy: A Hyperplane-Based Data Reconstruction Attack in Federated Learning, UAI 2025, pages 44–65

[6] Dwork and Roth, The Algorithmic Foundations of Differential Privacy, Foundations and Trends in Theoretical Computer Science 2014, pages 211–407

[7] Wei et al, Federated Learning With Differential Privacy: Algorithms and Performance Analysis, IEEE TIFS 2020, pages 3454–3469

[8] Sun et al, Soteria: Provable Defense against Privacy Leakage in Federated Learning from Representation Perspective, CVPR 2021, pages 9307–9315

[9] Jeter et al, OASIS: Offsetting Active Reconstruction Attacks in Federated Learning, IEEE ICDCS 2024, pages 1004–1015

[10] Scheliga et al, PRECODE: A Generic Model Extension to Prevent Deep Gradient Leakage, IEEE WACV 2022, pages 3605–3614

[11] Zhang et al, CENSOR: Defense Against Gradient Inversion via Orthogonal Subspace Bayesian Sampling, NDSS 2025

12- Building a compiler using generative artificial intelligence

Who?

Name : Sid Touati

Mail: Sid.Touati@inria.fr

Web page:

www-sop.inria.fr/members/Sid.Touati/

Where?

INRIA Sophia

Place of the project:

KAIROS team

https://team.inria.fr/kairos/

What?

Pre-requisites if any: Compilation, C programming

Detailed description:

Building compilers is a key activity in computer science. Its aim is to

translate a high level programming language to a machine code.

Traditionally, building a compiler is a hard task, and requires the

ability to use some useful tools like lex&yacc (or any other parser

generator tool).

With the advance of AI, some hard programming tasks can be done with

LLM. This project is a prospective, it is devoted to check if it is easy

of difficult to use an LLM to generate a compiler with lex&yacc. The

input language is a C dialect (subset of C) and the output is a

intermediate representation (a graph representation or a three address

code).

References:

- A. Aho, R. Sethi et J. Ullman. Compilateurs : Principes, techniques et

outils. Éditeur PEARSON. 2007.

- John R. Levine, Tony Mason et Doug Brown. Lex & yacc. Edition

O’Reilly. 1993

13- Optimizing Linux and hardware for Real-Time Network Workloads

The performance of computer workloads results from complex interactions between hardware, software, and operating system behavior. By default, operating systems such as Linux are designed to be general-purpose, balancing flexibility, performance, and energy efficiency. However, this generality often results in suboptimal performance. Significant speedups, particularly with NUMA architecture, can be achieved by carefully tuning kernel and hardware parameters to better match the specific characteristics of the workload. This issue becomes particularly critical when dealing with real-time network traffic, which demands high processing power, low latency, and high bandwidth.

This project will investigate performance optimization for real-time network processing, with a particular focus on workloads such as 5G RAN traffic and network security applications. The work will proceed in four main stages:

1. Benchmark Development – Design and implement a reproducible synthetic benchmark that models real-time network traffic.

2. Bottleneck Analysis– Build and apply performance analysis tools to identify system bottlenecks (e.g., CPU saturation, queueing delays, shared bus I/O constraints).

3. System Tuning – Derive and apply kernel and hardware parameter adjustments to mitigate the identified bottlenecks, and evaluate the resulting performance improvements compared to default configurations.

4. Multi-Workload Scenarios – Extend the study to environments where multiple workloads share resources under Kubernetes, assessing the trade-offs and optimization opportunities in such settings.

The project brings together benchmarking, system-level performance analysis, and applied optimization, offering concrete insights into how Linux environments can be adapted to support demanding real-time network workloads. Students will work with a variety of hardware platforms and operating systems. During the PER, efforts were primarily dedicated to developing the benchmark and identifying performance bottlenecks. This internship will build on that foundation by focusing on system tuning in complex scenarios and extending the approach to workloads beyond 5G (such as memory-intensive applications).

Who?

Name: Walid Dabbous

Mail: walid.dabbous@inria.fr

Web page: https://team.inria.fr/diana/team-members/walid-dabbous/

Name: Damien Saucez

Mail: damien.saucez@inria.fr

Web page: https://team.inria.fr/diana/team-members/damien-saucez/

Where?

Place of the project: DIANA team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: DIANA

Web page: https://team.inria.fr/diana/

Pre-requisites if any:

Familiar with Linux, basic system performance knowledge, highly

motivated to work in a research environment and exited to tackle hard problems.

Description:

This internship requires a good understanding of the computer and Linux kernel architectures.

Useful Information/Bibliography:

The Linux kernel

https://linux-kernel-labs.github.io/refs/heads/master/lectures/intro.html

https://doc.dpdk.org/guides/prog_guide/overview.html

https://lwn.net/Articles/675572/

https://cdrdv2-public.intel.com/321211/pci-sig-sr-iov-primer-sr-iov-technology-paper.pdf

14- Evolution over time of social networks

What?
Title: Evolution over time of social networks

Who?

Name: Frédéric Giroire, Sayf Halmi, Nicolas Nisse

Mail: Nicolas.nisse@inria.fr

Web page: https://www-sop.inria.fr/members/Nicolas.Nisse/

Where?

Place of the project: COATI team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: COATI

Web page: https://team.inria.fr/coati/

Pre-requisites if any: basic knowledge in graphs, algorithms and Python

Description: The goal of the project is to develop methods to analyse the evolution over time of a social network. We will consider as example the graph of scientific collaborations as we already have collected data from SCOPUS. The project will have two phases:

- Data analysis: In the first phase, the student will use our data set to study different metrics describing the evolution of the productivity of scientists. It will be interesting to compare the obtained results depending on the level of multidiscipinary of the scientist, their domain of research…

- Then, the student will compare the experimental results together with models for generating random graphs, namely models of Barabasi-Albert kind (preferential attachment models) with different functions of preferential attachements. This relies on our on going work where we have considered piecewise linear functions and studied the characteristics of obtained graphs (e.g. degree distribution, evolution of the degree of the nodes…)

Useful Information/Bibliography:

- [GNOST 23] Frédéric Giroire, Nicolas Nisse, Kostiantyn Ohulchanskyi, Malgorzata Sulkowska, Thibaud Trolliet: Preferential Attachment Hypergraph with Vertex Deactivation. MASCOTS 2023: 1-8

- [GNTS22] Frédéric Giroire , Nicolas Nisse , Thibaud Trolliet , Malgorzata Sulkowska : Preferential attachment hypergraph with high modularity. Netw. Sci. 10(4): 400-429 (2022)

- [Trolliet21] Thibaud Trolliet: Study of the properties and modeling of complex social graphs. (Étude des propriétés et modélisation de graphes sociaux complexes). University of Côte d'Azur, Nice, France, 2021

- [BA99] A.L. Barabasi and R. Albert. Emergence of scaling in random networks. Science, 286(5439):509–512, 1999.

15- Study and Experimental Evaluation of RIS-Assisted Localization Techniques in 5G Open-Source Platforms

Accurate positioning in 5G networks is a key enabler for emerging use cases such as industrial IoT, autonomous robotics, and smart logistics.

Reconfigurable Intelligent Surfaces (RIS) can enhance localization by manipulating radio propagation, especially in non-line-of-sight conditions.

Recent 3GPP releases have introduced advanced positioning features and metrics, including PRS, SRS, and angle-based techniques.

Open-source 5G platforms such as OpenAirInterface (OAI) and SRS RAN now enable practical experimentation with real 5G signals.

This project aligns with ongoing research on integrating RIS into 5G-Advanced localization frameworks.

It aims to bridge the gap between theoretical studies and experimental validation of RIS-assisted localization techniques.

The project’s objective is to investigate and experimentally validate RIS-assisted localization in 5G New Radio environments.

It will analyze how RIS configuration affects signal measurements such as ToA, AoA, and RSS for UE positioning accuracy.

The student will implement a controlled testbed using open-source 5G stacks to reproduce RIS-enhanced propagation scenarios.

Simulation and/or small-scale experimentation will be conducted to quantify positioning gains with and without RIS.

The study will also assess practical challenges such as synchronization, calibration, and channel estimation.

Results will support future work on RIS-aided localization standardization and performance modeling.

The student will:

• Review 3GPP positioning architecture and RIS-assisted techniques from the literature.

• Set up and configure a 5G NR test environment using OAI or SRS RAN.

• Integrate a simulated or hardware RIS model in the setup (e.g., via channel emulation).

• Implement basic localization algorithms (ToA/AoA-based) with and without RIS reflection paths.

• Collect and analyze positioning data to quantify improvement metrics.

• Document findings and propose optimization strategies for RIS-assisted localization in 5G NR.

The PER was dedicated to gaining hands-on familiarity with the 5G open-source platforms used in the project such as OAI or SRS RAN and to establishing the first elements of the experimental chain, including basic RIS integration and preliminary measurements of positioning indicators (e.g., ToA, AoA, RSS). This preparatory step provides the technical groundwork required to explore RIS-assisted localization in a controlled environment. The internship will build directly on these foundations by extending the testbed, implementing and refining localization algorithms, conducting systematic evaluations with and without RIS, and addressing practical challenges such as synchronization, calibration, and channel estimation.

Who?

Name: Walid Dabbous

Mail: walid.dabbous@inria.fr

Web page: https://team.inria.fr/diana/team-members/walid-dabbous/

Name: Damien Saucez

Mail: damien.saucez@inria.fr

Web page: https://team.inria.fr/diana/team-members/damien-saucez/

Name: Stefano Lioce

Mail Stefano.Lioce@inria.fr

Web page: https://scholar.google.com/citations?user=j7NF1AIAAAAJ&hl=it

Where?

Place of the project: DIANA team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: DIANA

Web page: https://team.inria.fr/diana/

Pre-requisites (if any)

• Strong background in wireless communications and 5G NR architecture.

• Familiarity with software-defined radios (e.g., USRP) and Linux-based networking.

• Basic programming and scripting skills (Python, MATLAB, or C++).

• Understanding of localization principles (TDoA, AoA, RSS).

• Prior exposure to OAI or SRS RAN software environments is desirable.

References

• Abuyaghi, M., Si-Mohammed, S., Shaker, G., & Rosenberg, C. (2024). Positioning in 5G Networks: Emerging Techniques, Use Cases, and Challenges. University of Waterloo & University of Strasbourg. https://hal.science/hal-04948666v1/document

• Mogyorósi, M., Papp, Z., & Vida, R. (2022). Positioning in 5G and 6G Networks—A Survey. Sensors, 22(13): 4919. https://pmc.ncbi.nlm.nih.gov/articles/PMC9268850

• Xanthos, S., Kalogirou, I., & Chatzinotas, S. (2022). RIS-Aided Joint Localization and Beamforming for Reconfigurable Intelligent Surface Aided 5G mmWave Communication Systems. arXiv preprint arXiv:2210.17530. https://arxiv.org/abs/2210.17530

Open-Source 5G Platforms

• OpenAirInterface (OAI) — https://openairinterface.org Open-source 5G RAN and Core Network software enabling real-world experimentation.

• srsRAN (Software Radio Systems) — https://www.srslte.com Open-source LTE/5G software suite for RAN development, testing, and research.

16- Logical Clocks as Infinite Sequences for Event-B and Rodin

Who?

Name: Frederic MALLET, Robert de Simone

Mail:Frederic,Mallet@univ-cotedazur.fr, Robert.De_Simone@inria.fr

Telephone:

Web page: https://www.i3s.unice.fr/~fmallet/

Where?

Place of the project:Kairos Team, Lagrange, Inria

Address: 2004 route des Lucioles

Team: Kairos (i3s/Inria)

Web page: https://team.inria.fr/kairos/

What?

Pre-requisites if any: Temporal Logics, Logical time, synchronous paradigm

Detailed description:

The work should be done in the context of the ANR PRC Project TAPAS (https://frederic-mallet.github.io/anr-tapas/).

TAPAS' main objective is to introduce a theory of time dedicated to Event-B and to combine model-checking with theorem proving in a unified framework. TAPAS should embedd several models of time, including a logical model of time inspired from the Synchronous languages like Esterel.

The actual work should study the environment of Event-B with Rodin and Pro-B, several instances of synchronous languages using logical clocks (CCSL, Esterel) and define a Event-B theory for logical clocks. This theory will rely on the theory of sequence that is already defined in Rodin. It should be used to prove (in a mechanical way) the properties of CCSL.

The outcome will be an open-source theory for Rodin that would allow to model classical CCSL requirement with Rodin and simple concurrent statements that would come from a multi-clock version of Esterel language. The work should be published at a venue like ABZ or ICFEM.

References:

F. Boulanger, C. Jacquet, C. Hardebolle, and I. Prodan, “TESL: a language for reconciling heterogeneous execution traces,” in MEMOCODE2014, pp. 114–123, ACM/IEEE, 2014.
J. Abrial, Modeling in Event-B - System and Software Engineering. Cambridge University Press, 2010.
M. J. Butler and I. Maamria, “Practical theory extension in event-b,” in Theories of Programming and Formal Methods, vol. 8051 of LNCS, pp. 67–81, Springer, 2013.
M. J. Butler, “Decomposition structures for event-b,” in Integrated Formal Methods, 7th Int. Conf., IFM, vol. 5423 of LNCS, pp. 20–38, Springer, 2009.
D. Cansell, D. M´ery, and J. Rehm, “Time constraint patterns for Event-B development,” in B2007, vol. 4355 of LNCS, pp. 140–154, Springer, 2007.
J. Abrial, M. J. Butler, S. Hallerstede, T. S. Hoang, F. Mehta, and L. Voisin, “Rodin: an open toolset for modelling and reasoning in event-b,” Int. J. Softw. Tools Technol. Transf., vol. 12, no. 6, pp. 447–466, 2010.
F. Mallet, C. Andr´e, and R. de Simone, “CCSL: specifying clock constraints with UML/Marte,” Innov. Syst. Softw. Eng., vol. 4, no. 3, pp. 309–314, 2008.
P. Le Guernic, J. Talpin, and J. L. Lann, “POLYCHRONY for system design,” J. Circuits Syst. Comput., vol. 12, no. 3, pp. 261–304, 2003.
H. N. Van, F. Boulanger, and B. Wolff, “A formal development of a polychronous polytimed coordination language,” Archive of Formal Proofs, 2019. http://isaafp.org/entries/TESL Language.html.
R. Alur, T. A. Henzinger, and M. Y. Vardi, “Parametric real-time reasoning,” in STOC, 1993.
E. Andr´e, “IMITATOR 3: Synthesis of timing parameters beyond decidability,” in CAV, vol. 12759 of LNCS, pp. 1–14, 2021.
A. Benveniste, P. Caspi, S. A. Edwards, N. Halbwachs, P. Le Guernic, and R. de Simone, “The synchronous languages 12 years later,” Proceedings of the IEEE, vol. 91, no. 1, pp. 64–83, 2003.
J. Colaco, B. Pagano, and M. Pouzet, “SCADE 6: A formal language for embedded critical software development,” in TASE 2017, pp. 1–11, IEEE, 2017.
L. Lamport, “Time, clocks, and the ordering of events in a distributed system,” Commun. ACM, vol. 21, no. 7, pp. 558–565, 1978.
J. Deantoni and F. Mallet, “Timesquare: Treat your models with logical time,” in TOOLS, vol. 7304 of LNCS, pp. 34–41, 2012.
M. Montin and M. Pantel, “Towards multi-layered temporal models: - A proposal to integrate instant refinement in CCSL,” in FORTE 2021, vol. 12719 of LNCS, pp. 120–137, 2021.
S. A. Edwards and E. A. Lee, “The semantics and execution of a synchronous block-diagram language,” Sci. Comput. Program., vol. 48, no. 1, pp. 21–42, 2003.
M. B. Dwyer, G. S. Avrunin, and J. C. Corbett, “Patterns in property specifications for finite-state verification,” in ACM ICSE’ 99, pp. 411–420, 1999.
Lionel Rieg, Gérard Berry: Towards a Coq-verified Chain of Esterel Semantics. Leibniz Trans. Embed. Syst. 10(1): 2:1-2:54 (2025)

17- Blockchain-backed Operational Transformation for Collaborative Editing

Who?

Name: Pascal URSO

Mail: pascal.urso@univ-cotedazur.fr

Where?

Place of the project: Laboratoire d'Informatique, Signaux et Systèmes de Sophia Antipolis (I3S)

Address: Euclide B, 2000 Rte des Lucioles

Team: SCALE

Web page: https://scale.i3s.unice.fr/

Pre-requisites: Good knowledge of distributed systems, good programming skills (e.g. Java, JavaScript/TypeScript, or Go), basic familiarity with blockchain / smart-contract or permissioned ledger technologies.

Goal / Description:

Collaborative editing tools — such as shared documents, code editors, real-time note-taking platforms, and distributed version control software system — are central to distributed work and research. Ensuring that multiple users can concurrently modify a shared document without conflicts requires sophisticated consistency management, traditionally achieved through quite complex or costly Operational Transformation (OT) or Conflict-free Replicated DataTypes (CRDT) algorithms. At the same time, blockchains and distributed ledger technologies offer tamper-resistant, decentralized storage with strong ordering guarantees.

The aim of this internship is to explore whether collaborative editing algorithms can improve their efficiency or capabilities by running over a blockchain (or permissioned ledger) substrate.

The student will:

Review OT/CRDT theory (convergence, transformation, tombstone-based approaches) and related work on ledger- or blockchain-based collaborative editing;
Propose a design mapping OT/CRDT history + transforms onto a ledger (handling ordering, forks/re-orgs, confirmations);
Implement a prototype: a simple collaborative editor whose operations are committed onto a blockchain / ledger (e.g. via smart contract or chaincode), and which applies OT transforms/CRDT computation based on the ledger history;
Evaluate the system (latency, throughput, correctness under concurrent edits or simulated ledger re-ordering), and compare with other collaborative editings algorithms baseline.

Expected Outcome: A working prototype + evaluation results + a written report (potentially a short workshop/conference paper) describing findings, trade-offs, limitations — plus source code as baseline for future research.

Key References / Reading (to start):

Collaborative Text Editing with Eg-walker: Better, Faster, Smaller. J Gentle, M Kleppmann. Proceedings of the Twentieth European Conference on Computer Systems, 2025
Decentralised / blockchain-based collaborative editing project: CryptPad / ChainPad. CryptPad Team. Whitepaper, 2023.
Real differences between OT and CRDT in correctness and complexity for consistency maintenance in co-editors. Sun, David, et al. Proceedings of the ACM on Human-Computer Interaction 4.CSCW1 (2020): 1-30.
Evaluating crdts for real-time document editing. M Ahmed-Nacer, CL Ignat, G Oster, HG Roh, P Urso. Proceedings of the 11th ACM symposium on Document engineering, 103-112
Asynchronous reconciliation based on operational transformation for P2P collaborative environments. M. Cart and J. Ferrié. In Proceedings of the International Conference on Collaborative Computing (CollaborateCom’07).

18- Conflict-Aware Editing Synchronization Between Humans and AI Agents

Who?

Name: Pascal URSO

Mail: pascal.urso@univ-cotedazur.fr

Where?

Place of the project: Laboratoire d'Informatique, Signaux et Systèmes de Sophia Antipolis (I3S)

Address: Euclide B, 2000 Rte des Lucioles

Team: SCALE

Web page: https://scale.i3s.unice.fr/

Pre-requisites: python/javascript programming, ideally with libraries like Yjs, or Automerge, or custom distributed algorithm implementations. Familiarity with LLM APIs (OpenAI, Anthropic, etc.) for multi-agent interaction. Interest in Git-based diffs / version history analysis.

Goal / Description:

With the rise of multi-agent LLM systems, collaborative document editing, code writing, and other content generation tasks increasingly rely on multiple AI and humans agents working in parallel. Traditional CRDTs (Conflict-free Replicated Data Types) ensure structural convergence for concurrent updates, but they operate at a syntactic level (insert, delete, move). LLMs operating with a global point-of-view and low awarness of collaborative partners often make semantically overlapping or conflicting edits, which syntactic CRDTs cannot detect. For example, two agents might rewrite a function completely differently, introducing contradictory statements. Structural convergence is guaranteed, but semantic coherence is not.

This internship explores how to augment CRDTs with more semantic awareness, enabling multi-LLM collaboration while maintaining both structural and semantic consistency.

Objectives:

Literature Survey: Study CRDTs in collaborative systems. Review recent multi-agent LLM frameworks, semantic-aware CRDTs (if any), and multi-LLM collaboration paradigms. Identify limitations of classical CRDTs in semantic-sensitive editing.
Study the difference of how LLMs and humans edit code.
Design a Semantic-Aware CRDT (S-CRDT) Model: Define a CRDT-like structure that captures semantic relations between document elements (paragraphs, sentences, code blocks). Integrate mechanisms for detecting semantic conflicts (e.g., contradiction, redundancy, style mismatch). Explore semantic resolution strategies: agent voting, priority rules, or automatic merging with LLM assistance.
Prototype Implementation: Implement a prototype S-CRDT for text documents (or optionally, code notebooks). Connect the system to a multi-LLM setup where agents make edits concurrently. Integrate semantic-checking via embeddings, similarity metrics, or LLM-based evaluations.
Evaluation: Test convergence: does the system still guarantee CRDT-style convergence? Test semantic coherence: quantify reductions in contradictions or inconsistencies. Compare to classical CRDTs in multi-agent editing scenarios.

Key References / Reading (to start):

Evaluating crdts for real-time document editing. M Ahmed-Nacer, CL Ignat, G Oster, HG Roh, P Urso. Proceedings of the 11th ACM symposium on Document engineering, 103-112
CodeCRDT: Observation-Driven Coordination for Multi-Agent LLM Code Generation. Pugachev, S. (2025). arXiv preprint arXiv:2510.18893.
The art of the fugue: Minimizing interleaving in collaborative text editing. Weidner, M., & Kleppmann, M. (2025). IEEE Transactions on Parallel and Distributed Systems.
Revisiting the Conflict-Resolving Problem from a Semantic Perspective. Dong, J., Sun, J., Lin, Y., Zhang, Y., Ma, M., Dong, J. S., & Hao, D. (2024, October). In Proceedings of the 39th IEEE/ACM International Conference on Automated Software Engineering (pp. 141-152).
Human-Written vs. AI-Generated Code: A Large-Scale Study of Defects, Vulnerabilities, and Complexity. Cotroneo, D., Improta, C., & Liguori, P. (2025). arXiv e-prints, arXiv-2508.

19- Exploring eBPF benefits in Kubernetes based clouds

cni-cilium.pdf

20- Privacy in Split Learning

---

# Who?

**Name:** Giovanni Neglia, Jingye Wang

**Mail:** giovanni.neglia@inria.fr, jingye.wang@inria.fr

**Web page:**

[http://www-sop.inria.fr/members/Giovanni.Neglia/](http://www-sop.inria.fr/members/Giovanni.Neglia/)

---

# Where?

**Place of the project:** Inria

**Address:** 2004 route des Lucioles, 06902 Sophia Antipolis

**Team:** NEO team

**Web page:** [https://team.inria.fr/neo/](https://team.inria.fr/neo/)

**Pre-requisites if any:**

The ideal candidate should have strong programming skills (Python,

PyTorch/TensorFlow).

An interest in machine learning and data privacy would be a plus.

---

# Description

Split Learning (SL) is a distributed training paradigm where a model is

split between a client and a server.

The client executes the first layers and sends the resulting activations

(the “smashed data”) to the server,

which completes the forward pass and backpropagation, returning

gradients to the client.

SL was introduced in [1] as a way to protect privacy by avoiding the

transfer of raw data.

However, recent works show that intermediate activations may still leak

information, enabling reconstruction of client data [2,3].

In this short preparatory project, we want to explore a simple idea:

**what if the client does not reveal the weights of its local head?**

If the server cannot align with the client’s architecture,

reconstruction attacks may become less effective.

The student will:

1. Survey briefly the main privacy attacks on split learning.

2. Set up a split learning pipeline (e.g., on MNIST or CIFAR-10).

3. Reproduce reconstruction attacks from the literature.

4. Test hidden-head variants where the client’s local layers remain

private, and compare reconstruction quality with the baselines.

The goal is to evaluate whether keeping the client head private can

reduce leakage.

This internship could lead to an industrial thesis with Hivenet [5].

---

# Useful Information / Bibliography

[1] P. Vepakomma, O. Gupta, T. Swedish, R. Raskar. *Split learning for

health: Distributed deep learning without sharing raw patient data.*

arXiv:1812.00564, 2018.

[2] X. Xu et al. *A Stealthy Wrongdoer: Feature-Oriented Reconstruction

Attack against Split Learning (FORA).* CVPR 2024.

[3] X. Zhu et al. *Passive Inference Attacks on Split Learning via

Adversarial Representation Alignment.* NDSS 2025.

[4] P. Vepakomma et al. *NoPeek: Information leakage reduction to share

activations in distributed deep learning.* ICDMW 2020.

[5] Hivenet, https://www.hivenet.com

21- Evolution over time of social networks

Who?

Name: Frédéric Giroire, Sayf Halmi, Nicolas Nisse

Mail: Nicolas.nisse@inria.fr

Web page: https://www-sop.inria.fr/members/Nicolas.Nisse/

Where?

Place of the project: COATI team, Inria, Sophia Antipolis

Address: 2004 route des Lucioles

Team: COATI

Web page: https://team.inria.fr/coati/

Pre-requisites if any: basic knowledge in graphs, algorithms and Python

Useful Information/Bibliography:

- [GNOST 23] Frédéric Giroire, Nicolas Nisse, Kostiantyn Ohulchanskyi, Malgorzata Sulkowska, Thibaud Trolliet: Preferential Attachment Hypergraph with Vertex Deactivation. MASCOTS 2023: 1-8

- [GNTS22] Frédéric Giroire , Nicolas Nisse , Thibaud Trolliet , Malgorzata Sulkowska : Preferential attachment hypergraph with high modularity. Netw. Sci. 10(4): 400-429 (2022)

- [BA99] A.L. Barabasi and R. Albert. Emergence of scaling in random networks. Science, 286(5439):509–512, 1999.

22- IA Générative pour l'histopathologie

Who?

> Name: Descombes Xavier

> Mail: xavier.descombes@inria.fr

> Telephone: +33(0) 4 89 15 43 24

> Web page: https://www-sop.inria.fr/members/Xavier.Descombes/

> Where?

> Place of the project: I3S

> Address: 2000 route des lucioles 04903 sophia antipolis cedex

> Team: morpheme

> Web page:https://team.inria.fr/morpheme/

> What?

Pre-requisites if any: IA générative, Traitement des images

Detailed description:

L'objectif de ce stage est de développer une IA générative pour simuler des images d'histopathologie avec un marquage immunohistochimique (IHC) à partir du marquage standard HE. La technologie préconisée est une approche par modèle de diffusion. Un logiciel devra être fourni et validé sur une base de données obtenue dans le cadre d'une collaboration avec le CHU de Nice. Cette validation s'effectuera en lien avec les anatomopathologistes du CHU. La pathologie concernée est le cancer du rein.

23 - Smart wireless sensor network for autonomous bird detection

Who: Benoit Miramond

Where: LEAT

Sujet_stage_2025_Birds_songs.pdf

Page updated

Google Sites

Report abuse