Building production-grade AI infrastructure across C++, Python, and TypeScript.
C++ AI infrastructure engineer. Most work proprietary; two pieces public below.
omnidyon-raptor — Hierarchical RAG · C++20
Standalone C++20 library implementing the RAPTOR paper — Recursive Abstractive Processing for Tree-Organized Retrieval. Traditional RAG retrieves isolated text chunks, limiting holistic understanding. RAPTOR recursively embeds, clusters, and summarizes chunks into a multi-level tree, then retrieves across all abstraction levels in a single query — granular detail and high-level summaries together.
Built on FAISS, libcurl, SQLite, and std::. Zero custom-utility dependencies.
omnidyon-ralph — Autonomous Agent Loops · C++20
Standalone C++20 library implementing RALPH — an autonomous development loop controller for AI agents. Three-state circuit breaker (Closed / HalfOpen / Open) prevents runaway loops. Intelligent exit detection requires both completion indicators and an explicit exit signal before stopping. Response classification across completion, questions, errors, and progress.
Agent-agnostic via consumer-implemented ITaskRunner — no dependencies on any specific agent harness. Reusable across any agent stack.
These two libraries are pieces of a larger, proprietary AI infrastructure stack I've built in C++20 — covering orchestration, observability, evaluation, and supporting tooling. Architecture deep-dives available in technical interviews.
C++20 Python TypeScript · CMake FAISS SQLite · FastAPI Pydantic · clang-tidy cppcheck Semgrep · GoogleTest · GCP/GKE
