I'm an AI/backend engineer who understands a system by rebuilding it — a Redis clone, a Java compiler with x86 codegen, a stack-based VM. Right now that same instinct is going into MCP (Model Context Protocol) servers and agentic AI integrations for production clients.
- Building production MCP servers — most recently FastAPI + DuckDB + JWT auth, deployed for a financial-data platform
- Shipping AI-first applications: RAG pipelines, LangChain/LangGraph agent orchestration
- Open to collaborating on anything that requires going deep enough to rebuild it from scratch
| Project | What it is | Stack |
|---|---|---|
| VKDB | Redis-compatible in-memory store — WAL, replication, AOF persistence | Java |
| Dataflow | Self-hosted pipeline tool for building pipelines & querying S3-backed data | Python, DuckDB |
| Mini Java Compiler | Full compiler — lexer, parser, semantic analysis, x86 codegen | C |
| Interpreter | Stack-based interpreter / VM | Go |
Languages
AI & Agentic
Backend
Data & Storage
Cloud & Infra
Game Dev (long-term bet)
Ask me about MCP servers, database internals, or why I keep rebuilding things other people would just import.