ML/AI engineer and infrastructure builder with 12+ years in tech — from industrial automation and HPC systems to AI agent orchestration and modern dev tooling.
I build tools that make AI agents work together reliably: orchestrators, policy engines, test platforms, and the infrastructure to run them all.
-
Maestro — AI Agent Orchestrator for parallel coding agent coordination. Runs multiple AI coding agents (Claude Code, Codex, Aider) on different parts of a project with DAG-based task scheduling and isolated git worktrees.
-
Arbiter — Coding Agent Policy Engine written in Rust. An MCP server that routes tasks to the best coding agent using Decision Tree inference, 10 safety invariants, and historical performance tracking. 10,000+ decisions/sec.
-
spec-runner — Task automation from markdown specs via Claude CLI. Reads structured
tasks.mdfiles and executes them with automatic retries, multi-agent code review, and Git branch management.
- atp-platform — Agent Test Platform. Framework-agnostic platform for testing AI agents with declarative YAML test suites, multi-level evaluation (artifact checks → behavior analysis → LLM-as-judge), statistical analysis, and CI/CD integration.
- ai-agent-management-system — Self-hosted AI agent management system with RAG integration, dynamic agent creation from templates, and Kubernetes deployment with HPA autoscaling.
- AI_guide_assistant — AI-powered guide assistant
- collab-editor — Collaborative editor (TypeScript)
- MD-Editor — Markdown editor (JavaScript)
- minitorrent — Minimal torrent client (Python)
| Domain | Technologies |
|---|---|
| Languages | Python, Rust, TypeScript, Bash, Elixir |
| AI/ML | LLM orchestration, RAG, MCP protocol, multi-agent systems |
| Infrastructure | AWS, Kubernetes, Docker, Terraform, EasyBuild |
| Data | PostgreSQL, Redis, Neo4j, ChromaDB, Qdrant, SQLite |
| DevOps | Jenkins, GitHub Actions, CI/CD pipelines |
| HPC | Linux cluster administration, scientific software deployment, RHEL |
By day I work as an HPC systems administrator supporting computational research infrastructure for drug discovery, genomics, and molecular modeling. By night (and every free moment) I'm deep in AI agent orchestration — figuring out how to make multiple AI agents collaborate on real engineering tasks reliably and safely.
I care about clean system architecture, reproducible experiments, and well-documented tools. I believe the future of software engineering involves humans and AI agents working side by side, and I'm building the tooling to make that work.
- GitHub: @andrei-shtanakov