Jeremy London
Engineering leader. Systems builder.
I build teams and production systems around AI, security, and distributed infrastructure. My work sits between engineering leadership and hands-on platform development.
Current role
Director of Engineering, AI & Threat Analytics
Working in
AI platforms, security, distributed systems
Usually building
Operator tooling, eval loops, private inference
The repeatable part
Different systems, same operating pattern.
The domain changes: AI platforms, agents, security, distributed systems, engineering teams. The useful work usually starts by making the blurry parts explicit.
01
Define the boundary
Name the owner, the interface, the failure mode, and the decision the system is allowed to make.
02
Make quality visible
Use evals, traces, tests, review, and operational signals so trust is measured instead of argued.
03
Keep it operable
Build the path for the person debugging, extending, or owning it after the demo is over.
Current interests
The thread running through most of my work right now is controlled autonomy: AI agents with real permissions, LLM gateways that route work intentionally, local inference that keeps private workflows close, and developer tools that make quality easier to see.
boundary
I keep coming back to the same questions: where should the boundary live, what does the system know about itself, and how does somebody recover when the clever path fails?
practice
That shows up in MCP tools, eval loops, WebGPU experiments, Kubernetes infrastructure, home automation, and the unglamorous parts of making AI systems maintainable.
Away from the keyboard
The non-technical side stays separate from the lab. It gives the rest of the work something real to bounce against.
Playing guitar
Usually loud enough to reset the room.
Cooking complicated food
Best when the recipe has one more step than expected or emulates a michelin star technique.
Snowboarding
Useful for getting out of debugging mode.
Taking photos
Mostly light, framing, and patience on my Ricoh GR III. The rest is luck.
Start with the writing that explains the work
The blog is a working notebook for AI platforms, engineering quality, agent systems, and the small operational details that usually decide whether software is usable after the demo.
AI platforms in production
Routing, observability, evals, cost controls, and the parts of AI systems that have to survive real users.
Agent workflow notes
Tool boundaries, approvals, handoffs, verifier loops, and the habits that keep long-running agent work legible.
Engineering quality
Testing, forms, state, scripts, review, and the small operational choices that make software easier to trust.
Security and trust
Authorization, validation, dependencies, audit trails, and product decisions that reduce risk without theater.
Worth reading first
A few useful entry points before you open the full archive.
Looking for something specific?
The blog now has search, category filters, year filters, and sort controls for the full archive.