Staff Software Engineer · Polyglot Programmer · AI Workflow Architect

13+ years shipping scalable, high-performance systems at Tesla, Meta, Etsy, WarnerMedia, and Regions Bank — spanning automotive, social, e-commerce, streaming media, and fintech.

I work at the intersection of polyglot engineering and AI-leveraged development. My two strongest muscles:

1. Cross-language fluency — I pick the right language for the problem instead of forcing every nail into the same hammer. I move between TypeScript, Python, Go, Rust, and the JVM without friction, and I've shipped production systems in each.
2. Building AI workflows — I treat LLMs and agents as first-class engineering teammates. I design multi-agent systems, build custom MCP servers, wire up code-review/audit bots, and automate the boring 80% of engineering work so humans can focus on the architectural calls that matter.

If a task can be delegated to an agent, I delegate it. If it can't yet, I build the harness so next time it can.

Languages I reach for daily, and what I use each for:

My everyday language for full-stack work. Production experience with Node.js, Bun, Deno, and modern browser runtimes. I write strongly-typed APIs, design type-level invariants that catch bugs at compile time, and ship React/Next.js frontends at scale.

My language for AI/ML workflows, data pipelines, and automation. I use it daily for LangChain, LangGraph, custom agent frameworks, FastAPI services, and the orchestration glue that ties LLM workflows together.

For high-throughput backend services, CLIs, and infrastructure tooling. Where I want predictable performance, simple concurrency, and a small binary I can ship anywhere.

For systems-level work and performance-critical paths — embedded inference runtimes, custom MCP servers, and high-performance data tooling. I lean on Rust when correctness and memory safety actually pay back the borrow-checker tax.

JVM stack from my fintech and enterprise work at Regions Bank — Spring Boot services, hardened core-banking integrations, and JVM-tuned high-throughput systems.

The glue. CI/CD pipelines, dev environment automation, one-off data surgery, and AI-agent orchestration scripts.

PostgreSQL, MariaDB, query-plan tuning, schema design, and the kind of analytics-grade SQL that makes ORMs look quaint.

This is where the polyglot background pays off most. AI workflow engineering is a system-design problem — you're wiring together models, tools, retrieval, evaluation, and humans — and every layer wants a different language and a different mental model.

• 🧩 Multi-agent systems — Specialized agents (research, code, review, audit, coordinator) that coordinate via message passing, shared memory, and structured handoffs. I design the topology, the anti-drift constraints, and the human escalation points.
• 🔌 Custom MCP servers — Model Context Protocol servers that expose internal tooling, codebases, dashboards, and proprietary APIs to LLM-based agents safely and with scoped permissions.
• 🧪 Agent harnesses for code review & audit — Automated PR reviewers, security audit bots, design-review agents, and CI-integrated quality gates that catch what humans miss and free reviewers for judgment calls.
• 🔁 Retrieval & memory architectures — Vector stores, hybrid search, graph-based context retrieval, and persistent agent memory that survives conversations and projects.
• 📊 Evaluation pipelines — Eval harnesses, regression detection on prompt changes, and observability for non-deterministic systems.
• 🚀 LLM-powered product features — Generative AI features shipped to real users at Etsy and beyond, with the latency, cost, and safety guardrails production demands.

• Agents over scripts when the work is fuzzy. Scripts over agents when it's deterministic. Know the difference.
• Eval the system, not the model. Prompts drift, retrieval rots, and models change underneath you. Build the harness first.
• Cost and latency are product constraints, not afterthoughts. Cache aggressively, batch when you can, and pick the smallest model that hits the bar.
• Humans in the loop where it matters. Automation everywhere it doesn't.

• Tesla — Led simulation platform efforts powering autonomy development.
• Meta — Built microservices and mobile platforms at planetary scale.
• Etsy — Explored and shipped Generative AI features for millions of buyers and sellers.
• WarnerMedia — Drove massive-scale migrations across NBA.com, HBOMax.com, and Max.com.
• Regions Bank — Hardened core banking systems with the reliability and compliance fintech demands.

Each move was deliberate — different domains, different scale problems, different engineering cultures — and the pattern compounds.

If you're building at the intersection of innovation, scale, and AI-leveraged engineering — let's connect.