I design and build AI agent infrastructure. Not prompt engineering — the layer under it: memory systems, multi-agent orchestration, autonomous pipelines, credential management. The stuff that makes agents actually work.
Not a project. Not a one-off. An ongoing relationship where I act as your agent infrastructure lead — part-time, remote.
Design the stack: gateway, knowledge store, credential proxy, deployment pipelines. Choose models, tools, and orchestration patterns.
Orchestration, handoff protocols, role-based tool access. Agents that chain together without context loss.
Cron, systemd, background workers. Agents that run benchmarks, audit infra, and deploy while you sleep.
Behind-the-firewall agent gateways. Compliance specs, audit trails. Fintech, health, defense.
Teach your engineers the 10-pattern methodology. They don't need me forever — they need the patterns.
Code quality vs agent readiness. Which models actually work for tool calling. Daily data, not vibes.
Series A/B startups that need agent infrastructure but can't hire a full-time architect. I set up the stack, train the team, stay on retainer.
Regulated companies (fintech, health, defense) that need on-prem agent gateways with compliance specs and audit trails.
Venture studios running multiple portfolio companies — one architect across all of them, same patterns, shared pitfall registry.
Engineering teams adopting AI agents for the first time and hitting the wall: agents forget, make the same mistakes, need constant hand-holding. I fix the infrastructure so the agents actually work.
1. We talk. 2. I audit your stack or design from scratch. 3. I build the infrastructure: gateway, knowledge store, skills, pipelines. 4. I train your team on the 10 patterns. 5. I stay on retainer — fractional, ongoing.
Remote. US/UK timezones. Async-first — Slack, GitHub, Notion.
No pitch deck. Tell me what you're building and where agents are breaking.
Also on workswithagents.dev · GitHub