— Independent AI consultant Remote · Worldwide Est. 2023 Vol. 04

For builders who'd rather ship than learn the whole field

AI that gets
shipped ,
not just talked
about.

I'm Michael Shvets — an independent AI engineer who works with solo founders and small teams to design, build, and quietly maintain LLM-powered products that actually pay for themselves.

Book a 20-min call → See recent work

~24h reply · No sales call to get a price

Live · try a question claude-haiku-4-5

Thinking…

— 01 / Services

What I actually do for you.

Four ways to work together. All of them start with a 20-minute call to confirm we're a fit — and to tell you honestly if we're not.

Strategy calls

One-hour deep dives. You bring the messy problem; I bring a battle-tested map and a written follow-up.

1 hrAsync recapStack-agnostic

Hands-on build

I sit inside your repo for two to six weeks and ship the AI feature your team can maintain after I leave.

2–6 wksEmbeddedCode + docs

Custom agents

Internal automations, support agents, data extractors. Scoped tightly, evaluated honestly, deployed to your stack.

Fixed priceEvals includedSelf-hosted opt.

Team training

A two-day intensive that turns your engineers into LLM-fluent operators. Less theory, more reps.

On-site2 daysCurriculum included

— 02 / Process

Boring, on purpose.

AI work fails the same way every time: ambiguous scope, no evals, vendor lock-in nobody noticed. This is the antidote.

Discovery call

Twenty minutes. We figure out if the problem you described is actually the problem you have.

Week 0 · Free

Written proposal

A one-page scope: what we ship, what we don't, the price, and the risks I see. No surprises.

Week 0–1

Build sprint

I'm in your Slack and your repo. Daily progress notes. End-of-week demos. You stay in the loop without managing me.

Week 1–6

Handover

Documented code, an eval harness your team can run, and a 30-day follow-up so it doesn't rot the moment I leave.

Final week + 30d

— 03 / Selected work

Things I've shipped recently.

Names removed because most clients prefer it that way. Happy to share references on the call.

Series A · Fintech

Cut their LLM bill by 71% without losing accuracy.

A growth-stage fintech was bleeding $18k/month on a single GPT-4 endpoint. We rebuilt the eval harness, routed 80% of traffic to smaller models, and kept the top-line accuracy within 1.5 points.

−71%

monthly spend

1.5pt

accuracy delta

3 wks

to ship

Solo founder · SaaS

Shipped a working AI writing tool in 18 days.

A non-technical founder with a 12k-person waitlist. From empty repo to paying customers in under three weeks — built on a stack they could afford to maintain alone.

days to launch

12k

waitlist users

infra/month idle

10-person team · Health

Built an agent that closed 4 hires worth of work.

A clinical-ops team was drowning in intake forms. We deployed a focused agent with strict evals and a human-in-the-loop checkpoint. It now handles 70% of intake autonomously.

70%

auto-handled

4 FTEs

of work absorbed

6 wks

incl. evals

Indie hacker

Untangled six months of agent spaghetti.

A solo dev with a working-but-fragile LangChain monstrosity. We replaced it with 400 lines of plain Python, added evals, and made the whole thing debuggable on a Tuesday afternoon.

6→1

dependencies

400

lines, total

hidden state

— 04 / Engagements

Three ways in.

Plain prices, no sales calls to find them. Discounts available for pre-revenue founders and non-profits — just ask.

The call

from $450

A focused hour for one specific decision. Best when you know what you're stuck on.

60-minute strategy call
Written recap within 48 hours
Two weeks of async follow-up
No prep deck required

Book a call →

Most picked

The build

from $12k

A fixed-scope sprint to ship one well-defined AI feature, end to end, with your team alongside.

2–6 week embedded sprint
Production code in your repo
Eval harness + docs handover
Weekly demos & async standups
30-day post-launch support

Start a project →

The retainer

from $6k

A few hours every week for teams that need senior AI judgment on tap without hiring full-time.

~8 hours/week, async-first
Slack + 1:1 each week
Code review on AI surfaces
Roadmap & hiring guidance
Month-to-month, cancel anytime

Inquire →

— 05 / FAQ

Questions people ask.

Didn't see yours? Send a one-line email. I read everything and reply within a day.

Who is this actually for?

Solo developers, founders, and small teams (think 2–20 people) who want to ship something real with LLMs without learning the whole field first. If you're a 500-person enterprise looking for a procurement-friendly vendor, I'm probably not your person.

What if my problem isn't really an AI problem?

Half the time it isn't. That's the kind of thing we figure out on the discovery call. If a Postgres query and a cron job will solve it, I'll tell you and refund your call fee.

Do you write code, or just advise?

Both. Strategy calls are advisory. Build engagements mean I'm in your repo writing TypeScript or Python alongside your team. I have strong opinions about boring code.

What models and stacks do you work with?

Mostly Claude, GPT, and open-weight models via vLLM or Bedrock. Comfortable with LangChain when it's the right answer and comfortable ripping it out when it isn't. I avoid frameworks that hide the prompts from you.

Can you sign an NDA?

Yes, before the discovery call. Standard mutual NDA, takes about an hour to turn around.

What if it doesn't work out?

Build engagements are milestone-based and you can stop after any milestone. The retainer is month-to-month. The strategy call has a no-questions refund.

AI that gets shipped , not just talked about.