Some of my time goes to early teams that need a principal engineer before they can hire one.

The parts of a young product that are cheap to get right now and expensive to change after the first enterprise customer.

• The platform under the product. The API surface, auth, the data model, and the queueing and storage underneath. I owned the REST API gateway at Twilio at 99.999% uptime, and operated the public GraphQL API and its storage at Weights & Biases. I've been a founding engineer twice: at Foam I built the authentication, authorization and flow-control middleware, and at DeepKeel I cofounded the company.
• Agents, and what they're allowed to do. At DeepKeel I built an LLM agent for cloud SaaS security, and a framework for inspecting and auditing the code LLMs write, so that an agent's actions can be validated and authorized before they run. At Ontologi I build OAuth-secured MCP servers that decide what an agent can read.
• Usage under load. Rate and concurrency limits for a multi-tenant API, which is one of my patents, and a queued write path at Twilio that held 99.999% success. Usage-based pricing is only as good as the counts underneath it.
• Knowing what it costs. At Material Security I led the Platform team and hired the TPM I set up a finops practice with. It took 30% off cloud spend in the first three months.

A set number of days a week, agreed up front, for a set term, with a review at the end of it where either of us can stop. I write things down as I go: design documents, decisions and the reasons for them. When you hire the team, they inherit a system that explains itself rather than one that only I understand.

I'm accountable for what I build and what I recommend, the same way I am in the audit. My background is on the homepage.