Embedded Files
Mohit Prabhushankar
Postdoctoral Research Fellow · Instructor of Record
Mohit Prabhushankar
Postdoctoral Research Fellow · Instructor of Record
School of Electrical and Computer Engineering
School of Electrical and Computer Engineering
Georgia Institute of Technology
Georgia Institute of Technology
Email: mohit.p@gatech.edu
I am seeking tenure-track faculty opportunities in machine learning, signal processing, and human–AI teaming, with a research program connecting foundational AI and human-collaborative systems!
About Me
About Me
I develop machine-learning methods that help people understand, question, and improve the behavior of trained models. Models are often trained under statistical assumptions, but deployed in open-world settings that allow interactions via prompts, retrieval, tool calls, memory, human feedback, and domain-expert interventions. These interactions expose a fundamental gap between the statistical training regime and the cognitive regime in which AI systems are used. My research develops the computational foundations and cognitive models for human-AI collaborative systems needed to close this gap.
I am a Postdoctoral Fellow in the Omni Lab for Intelligent Visual Engineering and Science (OLIVES) and an Instructor of Record in Electrical and Computer Engineering at Georgia Tech. I received my Ph.D. from Georgia Tech in 2021, advised by Ghassan AlRegib. My dissertation, Contrastive Reasoning in Neural Networks, studied how a model's response to alternative hypotheses can support explanation and reliable inference. My work has evolved from image-quality assessment and learned filter representations to foundation models and human–AI systems. Along the way, I have worked with natural images and video, ophthalmic imaging, seismic volumes, wearable and physiological signals, and, more recently, student–AI interactions.
Recent Updates
Recent Updates
Oct 11, 2026: Our work on active prompting titled “BALD-SAM: Disagreement-Based Active Prompting in Interactive Segmentation” is accepted at IEEE Access! [PDF]
Sept 25, 2026: Our work on Inferential Theory of Learning titled "Inferential Theory of Learning as a Framework for Test-Time Computation in Foundation Models" has been accepted at NeurIPS 2026!
Sept 15, 2026: Two coauthored contributions on visualization-first AI instruction and classroom accountability in Georgia Tech's Academic AI Playbook.
July 31, 2026: Our foundation-model robustness monograph titled "Robustness in Foundation Models: A priori, A posteriori, and Post hoc Robustness" has been accepted in Foundations and Trends in Computer Graphics and Vision!
Fall 2026: I am teaching Introduction to Signal Processing at Georgia Tech.
April 4, 2026: An in-progress website version for the Fundamentals of Machine Learning textbook is available here. Feel free to email me with suggestions and comments. This has been in the works for over a year now and I am excited to use the interactive notebooks during classes!
March 11, 2026: A preprint of our paper titled “BALD-SAM: Disagreement-Based Active Prompting in Interactive Segmentation” is available on Arxiv! [PDF]
February 24, 2026: Our paper titled "Hierarchical and Multimodal Data for Daily Activity Understanding” is published at the Journal of Data-centric Machine Learning Research (DMLR)! [PDF] [Code] [Link]
December 10, 2025: Our paper titled "A Large-Scale Benchmark on Geological Fault Delineation Models: Domain Shift, Training Dynamics, Generalizability, Evaluation and Inferential Behavior” is published at IEEE Access! [PDF] [Code] [Link]
November 11, 2025: Our paper titled "Countering Multi-Modal Representation Collapse Through Rank-Targeted Fusion,” is published in the The IEEE/CVF Winter Conference on Applications of Computer Vision 2026 (WACV)! [PDF] [Code]
November 5, 2025: I delivered a talk on "To use, and how to use, multimodal data: Imaging, Health Records, and Language (and everything else)" at the NSF Workshop on Accessible Healthcare in Atlanta!
July 22, 2025: I delivered a talk titled "The Learning-Robustness Paradox in Foundation Models: The Good, theBad, and the Ugly" at Tulane University, Department of Computer Science, hosted by Prof. Aron Culotta!
June 24, 2025: Our paper titled "Subject Invariant Contrastive Learning for Human Activity Recognition,” is accepted at the 2025 IEEE 35th International Workshop on Machine Learning for Signal Processing (MLSP) to be held in Turkey! [PDF] [Code]
May 30, 2025: Our paper titled "Multi-Level and Multi-Modal Action Anticipation,” is accepted at the IEEE International Conference on Image Processing (ICIP) to be held in Alaska! [PDF] [Code] [Link]
April 9, 2025: I delivered a talk titled "The Many Faces of Uncertainty in Post-hoc Neural Network Analyses" at the University of Pennsylvania, Wharton School, Department of Statistics!
April 8, 2025: Honored to receive Georgia Tech ECE's 2025 Roger P. Webb Research Spotlight Award!
Mar 2, 2025: Our paper titled "HEX: Hierarchical Emergence Exploitation in Self-Supervised Algorithms,” is accepted for an oral presentation at the 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)! [PDF] [Code] [Video].
Feb 28, 2025: Prof. Ghassan AlRegib, and I delivered our tutorial titled 'Inferential Machine Learning: Towards Human-collaborative Foundational Models' at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) at Tucson, USA! [Website]
Feb 26, 2025: Prof. Ghassan AlRegib, Prof. Xiaoqian Wang, and I delivered our tutorial titled 'Inferential Machine Learning: Towards Human-collaborative Vision and Language Models' at the 39th Annual AAAI Conference on Artificial Intelligence at Philadelphia, USA! [Website]
Jan 3, 2025: Our manuscript titled 'A unified framework for evaluating robustness of Machine Learning Interpretability for Prospect Risking' has been accepted for publication at Geophysics!
Dec 17, 2024: Our paper titled 'Targeting Negative Flips in Active Learning Using Validation Sets' won the Best paper Runner-Up Award at IEEE BigData'24 held at Washington DC, USA! [PDF] [Code] [Video]
Short Bio
Short Bio
Mohit Prabhushankar is a Postdoctoral Fellow in the OLIVES Lab and an Instructor of Record in Electrical and Computer Engineering at Georgia Tech. His research connects signal processing with inferential machine learning, robust representations, uncertainty, and human–AI collaboration, with applications in healthcare, geoscience, autonomous perception, multimodal sensing, and education. He received his Ph.D. from Georgia Tech in 2021. His contributions include research datasets, tutorials and short courses, an interactive machine-learning textbook, and work on accountable AI use in classrooms. His research and teaching have been recognized with awards from IEEE ICIP, IEEE Big Data, and Georgia Tech.
Contact
Contact
I welcome invitations for research seminars, tutorials, and discussions on inferential machine learning, reliable foundation models, human–AI collaboration, and AI-enabled education. Contact me via email.
Page updated
Google Sites
Report abuse