At Vitalops.ai, I optimized LLM inference and coding-agent harnesses for Blackbox AI. My vLLM and SGLang work on NVIDIA Blackwell GPUs reached more than 2× benchmark speed, with accuracy checked through custom evaluation suites.
At Vitalops.ai, I also shipped five products for enterprise customers across Supply Chain and Finance, working across full-stack engineering, AI agents, RAG, and model training and inference. I contributed to open-source projects including vLLM-Metal, Openvibe, Datatune, and Opendesk.
At Autonomio, I authored and open-sourced Jako, a distributed hyperparameter optimization tool, and EasyEnergy, a tool for tracking carbon and energy use in ML experiments. At Saama, I published BioSimCSE at ACL’s LOUHI workshop and built tooling to automate matching across medical databases.
As a freelance Machine Learning Engineer at Minicog, I built TableQA, an open-source NL-to-SQL engine, and published a research paper on the approach. I also co-authored a chapter in Advanced Deep Learning for Engineers and Scientists.
