Industry/Work Experience  

Qualcomm Wireless R&D Internship 

 Worked on a Cellular-V2X Project, using tools from Wireless Communication to manage workload offloading from a mobile device to edge and/or cloud server through a wireless network in dynamic network and sever environments.

Deutsche Bank (Financial Analyst)

• I was involved in monitoring risk for European GPF clients which included monitoring conditions for margin overrides, keeping an eye on liquidity, EM, concentration parameters and building tools for systematic and ad-hoc risk exposures visibility. 

• I worked for the integration of risk and margin framework between BNPP and DB. Built models to cover similarities and differences between DB and BNPP's stress and margin frameworks. 

• I have also been part of the process of quoting margins for new clients sample books and new trades on existing clients to help increase our platform balances in a sustainable and risk cognizant manner.

Selected Projects

Offloading computational workloads on the mobile edge communication-compute fabric (Research Project - In collaboration with Intel)

Motivation: 

• Increasing demand of computationally intensive mobile applications, e.g., AR/VR, computational perception, etc. 

• Resource constrained mobile devices, e.g., limited computation power, battery capacity, etc. 

• Leverage mobile edge communication-compute fabric (Mobile device(s) + Wireless communication + Edge compute)

• Heterogeneous wireless channels, mobile devices with different computation capabilities and possible network congestion

Problem: How to optimize workload and system design to make the most of available resources and adapt to an evolving system with heterogeneous wireless channel qualities, workloads, user types, etc.

Predicting Labor Condition Approvals for H-1B visas (Machine Learning - Course Project)

We present various approaches to predicting decisions on H-1B visa case statuses. The H1B visa enables foreign workers to work in the United States and is an important mainstay of immigrant participation in the US economy. We analyze 3 million records of H1B petition data from 2011-2016 provided by the US Office of Foreign Labor Certification and train machine learning models to predict the case status of the visas.  We trained 7 different models on the data and report various metrics for each model, and find that XGBoost is the best performer for this challenging task. We envision that these models will simplify the decision making process for the H1B visa.

Modelling and Analysis of Hierarchical Systems (Analysis and design of networks - Course Project)

High Level objective: Our central aim in this project is to model and analyze service systems such as a healthcare system, or a system to screen and censor controversial content on social-media websites. Such systems usually consist of workers of varying levels of expertise. Moreover, incoming tasks (patients for healthcare, controversial content for social-media) will be of varying levels of difficulties as well, which maybe apriori unknown.

Co-existence of eMBB and URLLC in IAB enabled networks (Wireless comm. - Course Project)

Abstract: Integrated access backhaul (IAB) is a new age deployment strategy in 5G NR with its merits and has become quite prevalent because of its low deployment costs. As also, two major service classes Ultra-reliable low-latency communication (URLLC) and enhanced mobile broadband (eMBB) have emerged. In this work we study the coexistence of eMBB and URLLC on the same IAB network. Now, the problem is interesting because URLLC applications demand strict latency and reliability whereas eMBB services require high data rates. We make use of a puncturing based scheduling to study the affect on delay and throughput. In the context of delay, we take cues from queuing theory to understand the affect of URLLC traffic on eMBB. In the context of throughput, we rely on simulations to see the affect. In the simulation results for delay we see a contention for resources by the two types of traffic i.e., they negatively impact each other. Similarly in the simulation results for achievable rate, the nonlinear relation between achievable eMBB throughput and URLLC throughput is observed. 

Positions of Responsibility 

Core Team Member (Counselling Service, IIT Kanpur)

Leadership: 

• Led a team of over 250 volunteers for providing academic, financial, and emotional support to students 

• Co-led a team of 136 student guides and coordinated with numerous faculty members for the smooth conduction of 6-day long Orientation Program for 830 freshmen at IIT Kanpur

Initiatives: 

• Conceptualized and conducted 7+ weekly sessions aimed at strengthening teamwork and interaction 

• Conducted Personality Enhancement Classes to help 40+ students develop good communication skills, openness to new things, and to build self-confidence and proactiveness interpersonally 

• Organized and assessed mental health survey for the campus community to make them acceptive of others

• Planned and Implemented a session aimed to make a start between students and faculty members  

Mentoring: 

• Guided and supported 3 freshmen academically and emotionally in their first year at IIT Kanpur

• Helped students with the course ‘Calculus’ and ‘Linear algebra’ by Institute and hall level remedial classes