I'm Brett Gillett. I founded the practice in December 2016, ran it as a team for a stretch, and reverted to solo on purpose. The work is better when the person you brief is the person doing the build.
I've been deploying things on Amazon Web Services since well before there was a Bedrock — back when "the cloud" still needed scare quotes. Hundreds of customer deployments later, the pattern that holds up best is also the simplest: small, opinionated engagements with one senior practitioner doing the actual work.
Curious Orbit started in December 2016. For a while we were a team. The work was good, but the bigger we got, the more time I spent on the work about the work — and the less time on the actual problems customers were paying me to solve. So I shrunk it back down.
Today the practice is one person — me. I take on a small number of engagements at once, and I'm the one writing the code, drawing the architecture, and standing in front of the leadership team when the recommendation lands.
Early 2026 I rewired how I work. My own practice runs on agents now: they draft, build, and ship alongside me, against a memory of every decision I've made. I sell what I run. The infrastructure work is still here, but it now sits underneath GenAI work that's the heart of the practice.
Concretely: three-week Bedrock feasibility sprints, multi-agent supervisor architectures, retrieval over messy enterprise data, and the unglamorous AWS plumbing that makes any of it production-grade — IAM, observability, cost discipline, day-two operations.
Scoped, not retained. Most engagements are weeks, not months. I'd rather earn the next one than inherit the inertia of the last.
Built in your account. Code ships into your AWS environment and your repo. No consulting infrastructure to migrate off later.
Honest about fit. If Bedrock isn't right for the use case, or if I'm not the right person for the job, I'll say so on the first call.
Plain language. Memos, not decks. Specifics, not certifications. The fewer acronyms a recommendation needs, the more likely it's right.
I keep bonsai. Twelve trees, in Oakville. There's a useful overlap with the day job: a bonsai is small on purpose, and most of the work is deciding what to cut. Both reward patience and knowing when to leave a thing alone.
Send a paragraph about what you're trying to do. If we're a fit, the next step is usually a 30-minute scoping call. If not, I'll point you somewhere better.
