San Gabriel, CA
Senior Backend Platform Engineer with 7+ years architecting resilient Kubernetes platforms, AI-powered tools, and multi-cloud DevOps solutions. Delivered large-scale migrations that cut costs 65%, boosted developer experience 125%+, and enabled zero-downtime operations. Recently shipped production AI integrations at Phantom and open-sourced MCP/agentic contributions. Passionate about scalable, intelligent infrastructure and actively seeking new opportunities. Read more
MCP stack, libraries, gateways, and product PWAs — use the icons for GitHub, npm, or the live site when each exists.
Project
Open-source MCP server: multi-protocol API supergraph (OpenAPI/Swagger/Discovery, optional GraphQL/gRPC), vault memory and bulk ingest, audit/cache, and optional sandbox, scheduling, Slack, HITL, and enterprise search — https://docs.clawql.com.
Project
Full-stack festival-planning PWA with hybrid retrieval (SQLite + embeddings + reranking), OpenRouter LLM integration, Tavily fallback, and group coordination — Lighthouse 90+ and deployed via ClawQL to Cloudflare and GitHub APIs.
Project
TypeScript gRPC transport for MCP: protobuf payloads, unary and bidirectional streaming, grpc.health.v1 probes, mTLS, and Kubernetes-friendly deployment — extends the official MCP SDK beyond stdio and Streamable HTTP.
Project
Pluggable evolutionary-loop library: typed Executor / Evaluator / Wonder / Reflect phases, ontology convergence gates, Postgres-backed lineage, MCP-shaped hooks, and an optional background seed poller.
Project
Streamable HTTP MCP gateway (optional gRPC) for /mcp at the edge: Panguard on stdio, upstream clawql-mcp-http, optional JWT (JWKS / Bearer on HTTP and gRPC), Helm mcpProxy integration.
Modules 18–20: STRIDE for agentic AI, OWASP Agentic Top 10 mitigations, and a quarterly security review checklist—merged here with links to the canonical curriculum.
Modules 13–17 in depth: automated containment, IR and PICERL, GPU abuse controls, secure local development, and production hardening for agent stacks—with links to every module.
Modules 10–12: data classification and PII in logs, verifying model weights, and observability for agentic workloads—consolidated with outbound links to each original module.
Modules 7–9 combined: RBAC and mTLS with Istio, Kata vs gVisor sandbox tradeoffs, and MCP runtime protection—plus links to the full modules on docs.clawql.com.
I’m particularly interested in roles blending AI agents, LLM tooling, or MCP with production platform engineering (Kubernetes, observability, reliability). Reach out if that sounds like your team.
Agents & APIs
Shipped ClawQL, an open-source MCP server: agents search and execute over OpenAPI/Swagger/Google Discovery (optional GraphQL/gRPC upstreams) with lean responses; bundled multi-provider graphs; vault memory (memory_ingest / memory_recall), audit, and cache; plus optional sandbox, schedules, Slack notify, Label Studio HITL, Onyx search, and more — stdio, HTTP, and gRPC. See https://docs.clawql.com.
On-device ML & GenAI
Forked and extended the Android “gallery” sample to explore local inference and GenAI UX — practical experimentation beyond pure backend work.
Platform at scale
Most recently at Phantom I architected multi-region Kubernetes, Istio, and CockroachDB failover systems; chaos engineering and K6 load testing; developer-experience scores up 125%+ and 133% on key CoreDX measures; and a 65% infra cost reduction after a Lambda-to-K8s migration.
More repositories and activity on GitHub.