About
We build the AI nobody else will scope honestly.
Spacecadet is a small AI-native product studio in Raleigh, NC. We take the hard, owned, proprietary-data builds — and we hand them off as code your team can keep.
Origin
Why a studio, and why AI-native
Spacecadet was founded after we watched the same pattern play out at company after company: the AI products most businesses actually needed couldn't be bought, and most studios couldn't actually build them. Vendors sold "AI-powered" SaaS that didn't understand the customer's data. Agencies bolted chatbots onto products that needed deeper rethinking. The good work — the proprietary-data, compliance-bound, architecture-defining builds — kept getting deferred because nobody would scope it honestly.
We built Spacecadet to take that work. We chose a studio shape (not a product company) because every serious AI build is its own problem and benefits from a team that ships, learns, and hands off ownership. We chose AI-native (not general dev) because the discipline is different — evals before features, real data in week 1, owned models over rented APIs.
Beliefs
What we believe about building AI
Most AI integrations fail because the model is the easy part.
The hard parts are the data pipeline, the evals, the deployment story, and the UX that lets a non-engineer trust the output. We treat all four as first-class engineering, not as scope creep.
Owning your data beats renting an API.
Vendor SaaS rents you intelligence on someone else's terms — your latency, your privacy boundary, your roadmap. We design AI products that put your data inside your own walls, and that work when the vendor pivots.
A working POC beats a polished plan.
You learn more about an AI product from one week of real data than from a month of architecture review. We get a functional proof of concept in front of your team by week 4, and we use it to decide what to build next.
Evals are the product, not a checkpoint.
No AI feature ships without measurable evaluations. The eval suite is the contract for what "done" means — and the regression net every future change has to pass.
Team
The people behind the work
Richard Haar
Co-Founder & CTO
Full-stack engineer with 12 years building products at startups and Fortune 500s. Leads architecture and the engineering practice — eval design, data pipelines, deployment.
LinkedIn →Andrew Thomasson
Co-Founder & CFO
Former lead at a top-tier commercial insurance consultancy. Runs commercial strategy, scoping, and the part of every engagement where price meets reality.
LinkedIn →Tony Shaw
UI Design & AI Lead
Has shipped AI features used by millions of people. Owns the product surface: how the AI shows up in the UI and how users learn to trust it.
LinkedIn →Neil Lewis
Marketing Director
Growth strategist who has scaled brands from seed to Series C and beyond. Leads positioning, messaging, and the rest of how Spacecadet shows up in the world.
LinkedIn →
What we say no to
The standards we keep
The work we refuse is as load-bearing as the work we take. Here is what we won't do, framed honestly.
We don't take projects where the AI is decorative.
If your AI feature could be a heuristic in a sprint, we'll tell you. We build AI where the model is genuinely load-bearing — not a marketing checkbox.
We don't do fixed-scope on AI builds.
AI is research. We engage in time-boxed cycles with explicit go / no-go gates. If you need a fixed-price line-item RFP, we are the wrong studio.
We don't engage without real production data access.
In week 1, with the actual data the model will operate on. Without it we cannot honestly scope, estimate, or evaluate the build — and we will not pretend otherwise.
We don't ghost.
Every engagement ends with a wrap-up doc: what shipped, how it works, what to monitor, and what we would build next if we were you.
If this sounds like the right fit, let's talk.
Book a 30-minute discovery call. We'll be honest about whether your project is the right shape for us.
Book a discovery call