Co-founded machinity.ai, a multi-tenant agentic AI platform (an autonomous project manager living in team chat) serving paying B2B customers — I designed and built its agent orchestration, RAG/semantic-memory layer, MCP tool governance and multi-provider LLM routing end to end. Before that: AI transcription pipelines for noisy aviation radio and 7+ years of backend/systems work in TypeScript, Python, Rust and C/C++ on Linux.
| Project | What it is |
|---|---|
| paper-pilot | MCP server that gives any AI agent real academic research: 16 tools over 6 scholarly databases, full-text PDF evidence extraction, citation graphs, Zotero sync |
| MCPCC | Rust gcc/clang wrapper that turns any compiled C/C++ binary into a typed MCP tool — CLI options auto-extracted into JSON Schema, stdio MCP server emitted per executable |
| Tartarus-Harness | Self-hosted multi-agent LLM stack in Rust: WebSocket chat service, multi-provider LLM proxy, MCP adapter unifying any MCP servers behind one endpoint, terminal TUI |
| PitchCheck | Cross-platform desktop app running Meta's TRIBE v2 brain-encoding model with multi-model LLM orchestration and automatic local-GPU/cloud provisioning |
| CodetoPromptGenerator | Desktop tool that turns any codebase into optimized LLM prompts: Tree-sitter code maps + LLM-driven smart file selection |
| showagent | Go TUI to browse, resume, branch and cross-convert local Codex and Claude Code agent sessions |
Taking remote AI/LLM contract and full-time remote engagements — UTC+3, full CET overlap, 4+ hours daily overlap with US East Coast, start within 2 weeks.