Fractional · monthly
You are shipping the second one. The first still needs you.
The first product works. It has paying users, and an AI feature somewhere in it that has to keep being right — extraction, retrieval, something that reads what a customer uploaded and turns it into a number someone will act on.
Now you are building the next thing. The first one has not stopped needing you; it has just stopped being the interesting problem. And the part of it most likely to break quietly is the part fewest people can maintain.
That is a capacity problem, not a project. Hiring a senior AI engineer takes four months and a salary you do not want to commit to. A $15/hr generalist can build CRUD and cannot make the AI part reliable — which you may already know first-hand.
The retainer
$2,500/mo
I own the AI surface of your product. It works, it is monitored, and when it breaks I am the one who fixes it.
Roughly two days a week. Async. Month to month, no notice period. Larger surfaces and second products are a larger retainer — I will say so on the first call rather than in month three.
The short answer
Fractional AI engineering from Code and Proof starts at $2,500/mo and buys ownership of the AI surface of a live product — extraction, retrieval, evals, and the pipeline that runs them — at roughly two days a week, async, month to month with no notice period. It is the answer when the AI already works and the problem is that nobody has time to keep it working.
How it works
I own a surface, not a ticket queue
Extraction, retrieval, evals and the pipeline that runs them are mine. You do not write specs for me, and you do not review my approach before I start.
Async by default
Slack or Discord, in your workspace. No standups, no weekly status call. If something needs a call it gets a call, and it is usually fifteen minutes.
Roughly two days a week
Enough to hold a live surface and move it forward. Not enough to also build your frontend, which is why that is on the excluded list rather than discovered in month three.
It is monitored, so I find out first
The failure mode of an unowned AI feature is that a customer notices before you do. Evals run on a schedule and the pipeline reports its own gaps.
Month to month
No minimum term and no notice period. A retainer that needs a lock-in to survive is one where the work stopped being worth it and nobody said so.
You can read everything I do
Commits in your repo, decisions written down where your team can find them. If I disappear tomorrow the next engineer is not starting from a black box.
What is and is not included
Written down here rather than negotiated later. A retainer without an edge becomes a full-time job at part-time money, and the person who suffers for that is usually the client, because the work they actually hired for stops getting done.
Included:
- The extraction, retrieval or agent surface of a live product
- Evals — building them, running them, and telling you when a number moves
- The pipeline around it: queues, retries, idempotency, cost ceilings
- Model and provider decisions, including when to stop paying for a bigger one
- Incident response on the AI surface when something breaks
- Written decisions your next engineer can read
Not included:
- Frontend work outside the AI features themselves
- Building a new product from scratch — that is a project, quoted separately
- Model training or fine-tuning. I do not do this, and I will tell you who does.
- Anything requiring a US entity, a SOC 2 questionnaire, or a security review
- General overflow development. I am not a body shop.
If most of what you need is on the right-hand list, I am the wrong hire and a call will not change that.
If you would rather start smaller
Handing someone a live surface on the strength of a sales page is a lot to ask. The reliability audit is five business days at a fixed price, and it ends with a written list of what will break and what fixing it costs. Most retainers here start that way — you find out how I work on something bounded, and I find out what I would be taking on.
Common questions
- What does a fractional AI engineer cost?
- Fractional AI engineering from Code and Proof starts at $2,500/mo. That buys ownership of the AI surface of your product — extraction, retrieval, evals, and the pipeline that runs them — at roughly two days a week, async, month to month with no notice period.
- What is not included in the retainer?
- Frontend work outside the AI features, model training or fine-tuning, and anything requiring a US entity or a security review. Building a new product from scratch is also outside the retainer, but only because it is a separate offer: that is AI application development, from $15,000. The exclusions are written down rather than negotiated later, because a retainer without an edge becomes a full-time job at part-time money.