Engineering Leader | AI Systems Builder | Platform Architect
Engineering leader with 20+ years building cloud-native distributed systems, developer platforms, and cybersecurity products. Currently building open-source AI agent tooling, including persistent memory systems for AI-assisted development and multi-agent workflows with intelligent model routing.
Engineering leader with deep experience across cybersecurity platforms, identity and access management (IAM), and cloud-native distributed systems. My career spans Fortune 100 scale and startup environments, and I specialize in connecting strategy to execution across product, engineering, and security stakeholders.
Most recently I have been building open-source AI agent infrastructure: persistent memory systems backed by vector databases, multi-agent workflows with cost-optimized model routing, and local AI workstation tooling. I bring the same engineering rigor from enterprise platform work to the AI tooling space.
MCP servers, vector search, multi-agent orchestration, and local LLM infrastructure
Security-first architecture for IAM, credential handling, and PII governance
Agile + DevOps transformation, observability, incident response, DORA metrics
Persistent AI second brain: a 41-tool MCP server backed by PostgreSQL + pgvector that gives every AI tool one shared memory. Thoughts are stored as vector embeddings, so Claude, Cursor, Windsurf, and Copilot search by meaning rather than keywords, and each one can read what the others learned. Stdio and streamable HTTP transports. Local-first; you own the data, and it runs for cents a month.
A guarded, cost-aware gateway between your application and AI providers. Callers declare a tier, not a model, and the router resolves the cheapest capable model across Anthropic, OpenAI, and local Ollama. Routing is priced from a registry, so what a tier selects and what it bills at cannot drift apart. Cache-aware sessions pin the model for a conversation: a provider's cached prompt prefix is keyed to the model, so switching mid-conversation throws the cache away and re-bills it at full price. That is the trap that makes naive routing lose money. Retry and failover, token and cost telemetry, and prompt-injection defenses throughout.
A two-agent LangGraph workflow with 3-tier model routing and Open Brain integration. A Researcher searches memory with pgvector semantic search, a Router scores task complexity across four dimensions using free heuristics with no model call at all, then routes to the cheapest capable tier: a local model for free, Haiku for a tenth of a cent, Sonnet only when the work genuinely earns it. An Analyst synthesizes the answer. If a cloud call fails, it falls back to the local model and the workflow keeps running.
Interactive setup wizard for a local AI workstation: LLMs, image generation, and video generation on any NVIDIA GPU. One command installs and configures Ollama, LM Studio, ComfyUI, and FramePack. No cloud dependency after setup.
A shareable discipline for doing agent work you can trust: plan, validate the plan, execute, validate the execution. At each gate two independent checks run together, a fresh reviewer that re-derives the answer from scratch and never sees the author's reasoning, and a deterministic check that actually runs. The producer never grades its own homework. Ships as a self-contained kit: the method, the automation to enforce it, and an optional tamper-evident ledger that records every rejection.
Led engineering for a cybersecurity platform serving 9,000+ enterprise customers. Established delivery guardrails, reliability practices, and DORA-based measurement.
Built API-driven data platforms processing billions of records daily. Delivered $20M+ annual cost savings and protected $500M+ in revenue through secure data separation.
Delivered AI/ML-powered classification and anomaly detection for suspicious access identification. Focused on secure SDLC, audit readiness, and operational resilience.
I'm always interested in hearing about new projects and opportunities. Whether you have a question or just want to say hi, feel free to reach out!