Skip to content
Rex St. John

Get in touch

Portrait of Rex St. John

Forward Deployed AI Engineer
San Francisco

I build developer platforms and agentic systems.

  • Forward-Deployed Delivery
  • AI Engineering
  • Agentic Systems
  • Evaluation & Benchmarking
  • Production LLM Apps

Python · PyTorch · TypeScript · React · Next.js · Node · Claude API · OpenRouter · MCP · Claude Code · LangGraph · Postgres · Docker · Kubernetes · Vercel · Modal · Solana

Sections

Experience

  • VerafyFounding Engineer. Built and run the public API, the TypeScript SDK and MCP server, and the production LLM apps behind them, including Swarm Explorer, which routes a claim to 250+ models.
  • AnzaCore Engineering Developer Relations. Operator-facing work on the Agave validator client for Solana, turning validator feedback into engineering priorities.
  • SagaHead of Developer Relations and Ecosystem. Built the function from the first hire. Founded the Saga Innovator Program: 224 projects onboarded in under 12 months.
  • NVIDIADevRel Engineer, Edge Computing. Embedded with 50+ edge AI startups, taking MLOps and computer vision workloads into production on Fleet Command. Shipped 50+ Helm chart templates for self-serve deployment.
  • ArmDirector, Developer Relations and Ecosystem. Founded the Arm Innovator Program. Designed an Arm-based autonomous vehicle reference architecture with ADLINK.
  • IntelDevRel Engineer and Technical Evangelist. Built the AWS IoT, Google IoT and Azure IoT developer kits with sample code, and shaped Intel Joule from 50+ CTO interviews.

Highlights

  • Forward-deployed deliveryScoping and integration inside customer engineering teams, production rollout, runbooks, debugging someone else's stack, onboarding and enablement, executive and stakeholder communication
  • Developer platformAPI design and versioning, SDK development, MCP servers, developer documentation and reference implementations, sample code, self-serve onboarding, sandboxes, CI for developer releases
  • Production LLM applicationsMulti-model routing and fallbacks, cost and latency control, caching, rate limiting and metering, credit systems, prompt-injection defense
  • Agentic systemsAgent orchestration, tool use, multi-model juries and consensus, commit-and-reveal protocols, failure handling, harness engineering
  • Evaluation and measurementLLM-as-judge juries, eval harness design, scoring and calibration, regression suites, reproducibility and provenance, benchmark construction, red teaming
  • AI securityPrompt-injection defense with published benchmarks, untrusted-input normalization, tool-access locks for agent juries, commit-and-reveal voting, security fuzzing, red teaming
  • Languages and frontendPython, PyTorch, TypeScript, React, Next.js, Node.js, Swift, React Native, iOS, Android
  • Infrastructure and DevOpsVercel, GitHub Actions, Hetzner, Coolify, DigitalOcean, Docker, Neon, Postgres and Prisma, Inngest, Modal, Ansible, Helm, Kubernetes, validator operations

Selected writing

All writing