Vancouver · Waterloo · New York · elsewhere

Hargun Mujral

I build fast systems that make machine learning useful outside of the notebook.

I studied computer science and mathematics at the University of Waterloo. Lately, I’ve been thinking about how to make inference faster, cheaper and easier to serve at scale.

/ Recent experience

Algorithm Engineer — Hudson River Trading

Systems engineering and infrastructure supporting quantitative researchers.

Student Researcher — Waterloo AI Institute / Dept. of Computer Science

Research on compound AI serving and medical imaging, spanning scheduling optimizations, fracture detection, and prostate cancer imaging.

ML Engineer — Databricks Mosaic Research

Built asynchronous distributed GPU training tooling and notebook infrastructure used by thousands of researchers and engineers.

ML Platform Engineer — X/x.AI

Designed the AI-powered captions system for millions of daily videos, plus low-latency speech and LLM inference infrastructure.

ML Researcher — Stratum AI

Full résumé

/ Selected projects

01 3Brown1Blue A multimodal AI tutor that listens, reads emotion, and generates animated explanations in the spirit of 3Blue1Brown. 02 ScadLM Open-source agentic CAD generation: natural language to editable OpenSCAD geometry. 03 Medical image processing Research on fracture detection and prostate cancer imaging with RCNNs, ResNet, and diffusion MRI. 04 SoMa Ramulator Extending a state-of-the-art DRAM scheduler with realistic memory traces and simulation tooling. 05 XCare An X-ray analysis suite combining computer vision and retrieval-augmented chat for medical imaging workflows.
All repositories

/ A few things I keep returning to