Services
Build it, then keep it working — and one of them costs $2,500 to try.
Everything here is the same discipline applied to a different surface: keep the model where it is genuinely good, keep it out of the judgement, and make every output able to explain itself.
The two groups below are not two businesses. Building an AI product and keeping one alive are the same skill observed at different points, and the reason the builds are worth buying is that the person doing them has spent the last two years being paged by the ones that were not built this way.
If you are not sure which applies, the audit is the cheapest way to find out — and sometimes the answer is that you do not need me.
Build it
Something that does not exist yet, or exists as a prototype nobody trusts. Priced from a published floor rather than discovered on a call.
AI application development
An AI product built end to end — the pipeline, the evals, the review surface a person actually uses, and the operational spine that keeps it running six months later.
AI agent development
Agents that are allowed to act in production. A human gate before anything irreversible, idempotency on every side effect, cost ceilings enforced in the runtime, and a transcript you can read at 2am.
AI MVP development
A first version narrow enough to build honestly and real enough to put in front of users. Evals written before the feature, so version two is a measurement rather than an argument.
White-label AI development
You sold the AI feature; I build it under your brand. Contractually, I do not sell the surrounding work and do not approach your client.
Operate it, fix it, or find out
Something already running — or already wrong. This is the half that decides whether the build above was worth doing, which is why it is not a separate business.
Fractional AI engineering
A senior AI engineer who owns the AI surface of your product month to month — extraction, retrieval, evals and the pipeline that runs them. Roughly two days a week, async, from $2,500/mo.
AI reliability audit
Fixed scope, fixed price. I read your existing AI feature and write down what will break, why, and what fixing it costs. Most engagements start here.
AI project rescue
Your AI feature is in production and getting things wrong. I find out why, tell you what it costs to fix, and then fix it — evals first, so you can see the improvement rather than take my word for it.
AI integration
Putting AI into a stack that already has users. Idempotent side effects, real retry safety, enforced cost ceilings, and a human gate before anything irreversible.
RAG and retrieval
Retrieval with a golden set, recall measured separately from generation, and a regression gate in CI — so a prompt change that makes it worse fails the build.
Every published figure is collected on one pricing page, including how billing works and what is never charged for.