Engineer. Researcher. Athlete.
Computer Engineering @ TMU
Atish is a Computer Engineering student at Toronto Metropolitan University with a growing curiosity for how intelligent systems work, from the code that trains them to the silicon that runs them.
Outside of campus, he's spent nearly 8 years training Brazilian Jiu-Jitsu. He's a blue belt, a provincial silver medalist, and coaches the kids' class at his gym. It's shaped how he approaches most things: stay patient, keep learning, and trust the process.
Responsible AI Lab, Toronto Metropolitan University
Assisting with graduate Natural Language Processing (NLP) research in multi-hop question answering and knowledge distillation — an area focused on how complex reasoning can be compressed from large language models into smaller, more deployable ones. Contributions include running fine-tuning and distillation experiments, evaluating model performance across iterations, and supporting the broader research workflow as the project progresses. Full project details remain unpublished pending the graduate thesis defense.
Department of Physics, Toronto Metropolitan University
Contributed to a physics education research project spanning three introductory courses for engineering students: Mechanics, Waves and Fields, and Solid State Physics — studying how restructured lectures and engagement, tutorials and materials affect student learning outcomes.
The work touched every phase of the research cycle — from co-designing interventions grounded in retrieval practice, metacognitive learning, and collaborative problem-solving, to building and administering surveys across three course terms that collected over 700 student responses, to carrying out the analysis that measured their impact.
The findings supported two accepted conference presentations, as displayed below.
Publications
Rebello, C. M., Megally M., & Kabiraj, A. (Accepted, 2024) Employing a hybrid of problem-solving and retrieval prompts in introductory physics, contributed talk presented at the 2024 Summer Meeting of the American Association of Physics Teachers, July 6-10, 2024, Boston, MA, U.S.A.
Rebello, C. M., Megally, M., Kabiraj, A. (Accepted, 2025). Exploring Student Success in Undergraduate Physics Using a Hybrid of Problem-Solving and Retrieval Practice Prompts. Submitted paper to be presented at the 2025 Annual International Conference of the National Association of Research in Science Teaching, March 23-26, 2025, Washington, D.C., U.S.A.
A from-scratch implementation of a simplified systolic array — the class of hardware accelerator that underlies Google's Tensor Processing Unit. Built around a grid of Processing Elements (PEs), each performing multiply-accumulate (MAC) operations to execute matrix multiplication with minimal memory overhead.
The goal isn't to replicate production silicon. It's to understand, at the register level, what actually runs beneath the models.
View on GitHub →Whether it's research, collaboration, or just a conversation.
Or find me on GitHub → github.com/redrumspree