At Applied Research Institute, I built a multi-agent code-review gate with deepagents and LangGraph, integrated with self-hosted GitLab. Teams used it to catch code smells, design issues, and violations of team standards.
I designed the reviewer to assign focus areas such as security and performance to separate agents, each guided by a YAML prompt. The agents also read related files and merge request descriptions using agentic RAG.
At EntSpace, I worked on Strategist AI, a business-analysis product built with LangGraph and ChromaDB that shipped to clients in Docker. I also chose, tuned, and deployed the transcription and summarization components for Vido.
I built the core EdTech platform backend from scratch with FastAPI and PostgreSQL, including its deployment and CI/CD. In my doc-extract-eval project, I compared four models across three document-input modes and built a FastAPI extraction endpoint with caching and request limits.
