Technical Reviewer — Packt Publishing
Credited technical reviewer for OpenClaw AI in Production (Ken Huang): full code review of agent architectures, MCP integrations, and security boundaries against pre-publication releases.
Multi-agent systems on open protocols (MCP, A2A) · frontier-model evaluation · security-first harness engineering · Kubernetes · 15+ years of financial-risk data engineering.
Credited technical reviewer for OpenClaw AI in Production (Ken Huang): full code review of agent architectures, MCP integrations, and security boundaries against pre-publication releases.
Reproducible benchmark environments for frontier agentic models across ML, finance, and agentic domains — PyTest, Docker/Kubernetes, Terminal-Bench, Harbor.
Hardened VPS multi-agent systems with model routing, scoped permissions, and execution boundaries; RAG pipelines; deterministic containerized infrastructure.
Seven services: an orchestrator delegates over A2A to three LLM agents, each reaching its own MCP tool server (2026 stateless spec), with live OSV.dev intel and human-gated publication. The model is treated as an untrusted component — every guarantee lives in the harness, and every guarantee is a test.
Financial data ingestion and analytics for derivatives, crypto, and market data; QuantConnect LEAN backtesting; PostgreSQL and Linux.
Entity-resolution and watchlist-classification systems spanning OFAC, HM Treasury, EU, and UN lists; large-scale financial-risk pipelines for compliance workflows.