Most AI PMs can't read a LangGraph trace. Most engineers can't ship a product. I've spent 6 years doing both building multi-agent systems from architecture to production, and shipping 0→1 products in regulated environments where correctness isn't optional.

My edge: I understand where agentic systems break, tool call failures, memory drift, non-deterministic planning loops, MCP coordination overhead. That's not a skill you pick up reading papers. It comes from shipping.

6 years · Multi-agent AI · DeFi · RWA Tokenization · Agentic Infrastructure

Open source multi-agent library with tool-use, memory, and planning loops.

Built to replace manual operational workflows with autonomous AI execution. Handles real failure modes, retry logic, memory drift, tool call failures. Not a demo. Not a wrapper. A production system.

Stack: Python · LangGraph · CrewAI · Claude API · n8n · MCP
Status: Active · Accepting PRs

Anthropic's Claude Code terminal wasn't running natively on desktop. I fixed it.

Diagnosed OS-level path and config failures, refactored the build pipeline, shipped cross-platform. Community picked it up within weeks, proof that real users had the same problem.

Stack: Node.js · Electron · Anthropic SDK
Status: Active · Open source

Compliant fractional on-chain ownership system for real-world assets.

End-to-end delivery: architecture → legal → compliance → smart contracts → mainnet. Zero missed milestones across 6 months. This is what "PM with engineering depth" looks like in practice.

Stack: Solidity · Hardhat · Chainlink oracles · OpenZeppelin
Status: Live on mainnet

Agentic AI systems       ████████████░░  Building
MCP integrations         ████████░░░░░░  Exploring
Agent governance         ██████░░░░░░░░  Researching
RWA + AI convergence     ████████████░░  Shipping

Claude Opus 4.6 Claude Sonnet 4.6 GPT-5 Turbo Gemini 2.5 Pro Grok 4
Llama 4 Scout Llama 4 Maverick DeepSeek V4 Qwen 3 Mistral Medium 3

LangGraph LangChain CrewAI AutoGen v0.4 OpenAI Agents SDK
Anthropic Agent SDK Google ADK Mastra Semantic Kernel Pydantic AI

MCP (Model Context Protocol) A2A Protocol OpenAI Tool Use Function Calling
Computer Use Claude Code Gemini CLI

n8n LangFlow Dify Zapier AI Flowise

LlamaIndex Weaviate Pinecone Chroma LangSmith OpenTelemetry

Vertex AI Azure AI Studio AWS Bedrock Groq Together AI Fireworks AI

Solidity Hardhat Web3.js Hyperledger Fabric Corda OpenZeppelin Chainlink

JIRA Confluence Notion GitHub Actions Agile Linear

I write about what's breaking in production, not what's trending. 3x/week on LinkedIn.

Recent topics:

• What actually fails in agentic systems at scale — and how to design around it
• Why RWA tokenization is a 10-year infrastructure play, not a hype cycle
• The PM skills AI augments vs. the ones it exposes
• MCP is becoming the HTTP of agent-to-tool communication — here's what that means
• How blockchain engineering trained me to reason about agentic failure modes

→ Follow on LinkedIn for 3 posts/week