I like building the backend of AI systems more than the AI itself — the ingestion pipelines, the retrieval layer, the service that has to stay up and stay fast. Most of what I build sits at the intersection of Go for the systems side and Python/RAG for the ML side.
Right now I'm mostly thinking about repository intelligence, RAG infrastructure, and what it actually takes to run these pipelines reliably at some scale — not just get them working once in a notebook.
I also spend a fair amount of time reading papers and rebuilding the ideas myself, because that's usually the fastest way I actually understand how something works.
Currently learning: distributed systems, high-performance Go, AI infrastructure, search & retrieval systems.
KnowYourRepo — paste a GitHub URL, get back a call graph, dependency map, and a RAG-based chat interface for the codebase. Go API server, Python/FastAPI ML sidecar, Next.js frontend.
AI Architecture Diagram Generator — turns a plain-language description into an Excalidraw diagram. Runs as a small pipeline of agents (parse → layout → validate → repair → export) orchestrated with LangGraph, FastAPI backend, React/TypeScript frontend.
From Scratch —A personal repository where I build foundational machine learning, deep learning, and reinforcement learning concepts from the first principles— without relying on high-level libraries or shortcuts.
SmartFile Manager — a distributed semantic file search pipeline in Go + Python. First project I built that a senior dev actually reviewed and rated well.
Semantic-Duel — a real-time multiplayer word guessing game, Go + WebSockets + ML.
Building scalable systems that power intelligent applications.