Hentati Hentati

I'm a software engineer who builds AI-native systems end to end, and knows how to step into existing codebases, refactor them, and make them better.

At Carleton University in Canada, I worked on research-grade software, migrating legacy build systems and refactoring tools for automated testing. I learned that the hardest engineering isn't building from scratch, it's improving what already exists without breaking what works.

I spent months reading multilingual incident reports manually for enterprise automotive clients, so I built IncidentIQ, a RAG-based analysis platform using Gemini embeddings and cosine similarity that does it automatically. I was tired of writing tone-matched email replies by hand, so I built Vibe, an MCP server that reads a contact's full Gmail history via OAuth 2.0, detects how they actually write, and generates replies that sound exactly like them. It ships as a Chrome extension injected into Gmail, a FastAPI backend, and an MCP server with SSE transport.

My stack is Python, FastAPI, NestJS, TypeScript, React, PostgreSQL, and modern AI tooling, MCP, RAG pipelines, vector embeddings, Gmail API.