About

I'm a Staff AI Researcher at Huawei Canada in Vancouver, where I work on efficient inference and reinforcement learning post-training for large language and multimodal models. My research spans machine learning and distributed systems, with a focus on multimodal inference, elastic mixture-of-experts serving, and resource-efficient training.

I have worked on distributed inference and post-training for the Qwen, GLM, and DeepSeek model families, including Qwen’s 235B mixture-of-experts models and DeepSeek V3. My systems experience includes Ascend NPUs and the CloudMatrix 384 platform. My research includes Encoder–Prefill–Decode (EPD) disaggregation, an approach now adopted in major inference frameworks such as vLLM and SGLang.

I received my MSc in Computer Science from the University of British Columbia, where I was advised by Jim Little and Leonid Sigal. Previously, I was a Research Engineer at Aspiring Minds Research Labs, where I led an end-to-end ML product pipeline for automated assessment of computer programs, from problem formulation and dataset creation to model development and deployment. I was also a recipient of the KVPY fellowship, awarded by the Government of India to support students pursuing research.

My work has been published at ICML, CVPR, KDD, WACV, and BMVC, and I have filed 15+ patent applications related to machine learning and AI systems. I have served as a reviewer for conferences including NeurIPS, CVPR, ECCV, and AAAI.

Community: As a former ML-India coordinator, I founded the Chandigarh chapter, organized meetups, and interviewed researchers from academia and industry. I also helped introduce school students to data science.

Open source: I contribute to AReaL and develop open-source systems alongside my research. Earlier, I maintained Ubuntu’s touchpad-driver package and served on its Bug Triage team. At UBC, I helped build and maintain the computer vision lab website.

Outside of research, I enjoy playing pickleball and spending time in nature, especially hiking.