At BdREN Innovation (Cortex), I led the five-person crawler team for a system that crawls websites into a vector database and answers questions with citations. I built the CUET scraper in Python, capturing 735 documents, 382 faculty profiles, and 1,106 linked files with source provenance.
At Fleet AI, Inc., I authored more than 70 long-horizon agentic AI evaluation tasks across simulated enterprise environments and QA-reviewed peers’ tasks. At PIERRA, I shipped a bilingual English–French website for an exterior design firm using Next.js and Tailwind CSS.
I also built projects including Greenlight, a 3D robot arm simulator and IUT Techathon 2026 finalist, and DimSum, an LLM-assisted energy optimizer whose median model latency I reduced from 5.7 seconds to 2.3 seconds. My work also includes a personal finance app, Chrome extensions, and research on machine learning and explainable AI.
