I engineer deterministic agent runtimes, modular multi-agent systems, and tool-augmented LLM infrastructure.
My work focuses on transforming stochastic language models into reliable software systems by enforcing strict type boundaries, multi-loop orchestration (planning, execution, critique, and governance), and standardized Model Context Protocol (MCP) integrations.
- ⚡ Deterministic Agent Runtimes: Building structured execution engines with strict runtime validation (
Zod), bounded iteration loops, and real-time observability. - 🔁 Multi-Loop Orchestration: Designing interlocking loops for task planning, execution, self-reflection/critique, and human-in-the-loop governance.
- 🔌 Tool Protocols & MCP: Authoring Model Context Protocol (MCP) servers and proxies that bridge foundation models safely to filesystem, web, and CLI tools.
- 🤖 Automated Engineering Workflows: Developing self-correcting agents for automated codebase refactoring, pull request reviews, and issue triage.
┌──────────────────────────────────────────────┐
│ 🎯 REASONING & TASK INPUT │
└──────────────────────┬───────────────────────┘
│
▼
┌──────────────────────────────────────────────┐
│ 🦖 AGENTIC ORCHESTRATION LOOP │
│ (Planning • Execution • Reflection) │
└──────┬───────────────┬───────────────┬───────┘
│ │ │
┌─────────────────────┘ │ └─────────────────────┐
▼ ▼ ▼
┌─────────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ 🧠 State & Memory│ │📋 Task Decomposition│ │🧪 Self-Reflection│
│ (Vector/SQLite) │ │(Execution Trees) │ │ & Quality Gating │
└─────────────────┘ └──────────────────┘ └──────────────────┘
│ │ │
└─────────────────────┬───────────────┴───────────────┬─────────────────────┘
│ │
▼ ▼
┌─────────────────────┐ ┌─────────────────────┐
│🔌 MCP Tool Integration│ │🛡️ Safety & Governance│
│(APIs, Web & CLI Ops)│ │ (HITL & Policy Ops) │
└─────────────────────┘ └─────────────────────┘
| Layer | Technologies & Frameworks |
|---|---|
| Languages & Runtimes | TypeScript, Node.js, Python, PyTorch, C++ |
| Agent Architecture | Custom Monorepo Runtimes, LangGraph, AutoGen, Ollama |
| Protocols & Validation | Model Context Protocol (MCP), Zod Schemas, REST/WebSockets |
| Tooling & Storage | Vitest, Turborepo, Docker, SQLite, ChromaDB |
- 🦖 Frankenbeast — Modular deterministic agent runtime & Turborepo framework.
- 🐝 SmartSwarm — Multi-agent swarm orchestration platform.
- 🤖 hermes-agent — Autonomous review and workflow agent.
- 🔌 mcp-servers & mcp-web — Custom Model Context Protocol tool suite.
- 🛠️ codex-review — Reusable review loops for automated PR gating.
- 🧩 agent-skills — Custom skill templates for AI agent coding.
- ⚡ ollama-orchestrator — Lightweight local LLM orchestration.
⚡ Building deterministic agent infrastructure & modular LLM tooling