From kickoff to a live, evaluated agent or copilot you can put in front of real users.
// AI ADVISORY · WARSAW, PL
Apply AI where it actually moves the business.
Most AI projects stall in slide decks. I work shoulder-to-shoulder with founders and ops leads to ship the right AI use cases — agents, copilots, NL-to-data, internal tools — to production in weeks, not quarters.
What you walk away with.
Ranked by ROI, time-to-value, and risk — so you spend the budget on what compounds, not on demos.
On routine workflows we automate end-to-end (support triage, data Q&A, reporting, lead qualification).
I don't resell models, infra, or platforms. Recommendations are picked on merit, not commission.
I don't pitch AI from a slide deck.
SV Lab runs three live AI products in production today, on top of a 12-year track record of shipping internet products end-to-end. Below: the work most relevant to this engagement.
Prowl
prowl.chatOne MCP endpoint. 408 market-intelligence tools for your agents.
The research layer for coding agents. Connect one MCP to Cursor, Claude Code, Codex — or any MCP client — and the agent can call 408 SEO, ads, SERP, review, and market-data tools with real numbers, cross-referenced into a 12-module strategy report shipped as an interactive brief, infographic, PDF, deck, and video. Billed per call from a USD wallet, with a $5 starter credit.
Agent intelligence · MCP · 2025 → Present
Visit
CheckMyData
checkmydata.aiChatGPT for your database.
Open-source, MIT-licensed agent that turns plain-English questions into queries, executes them across PostgreSQL, MySQL, ClickHouse, and MongoDB, and explains results with auto-generated charts. Privacy-first by design and fully self-hostable.
Open-source · NL → SQL · 2025 → Present
Visit
How it works.
- 01 · WEEK 1
Discovery & opportunity map
We map your workflows, data, and tooling, then score 8–12 candidate AI use cases on impact, feasibility, and risk. You leave the week with a ranked roadmap.
- 02 · WEEKS 2–4
Working prototype
I build the top use case end-to-end — agent, eval harness, guardrails, and a thin UI — running on your data. Not a slide deck, not a demo: a tool your team can use.
- 03 · WEEK 5+
Hand-off or scale
Either I package the prototype into a shippable internal product, or I hand it to your engineering team with docs, evals, and an upgrade path. Your call.
Common questions.
Do you only work on greenfield projects?
No. Most engagements are inside existing companies, on top of existing data and tooling. I'm comfortable joining a stack rather than rebuilding it.
Which models / platforms do you recommend?
Whichever ones fit. I've shipped on OpenAI, Anthropic, and open-weight models, with and without orchestration frameworks. The choice is driven by latency, cost, privacy, and the eval bar — not by vendor preference.
What does engagement look like?
Either a fixed-scope advisory sprint (1–4 weeks) or a fractional retainer (1–2 days/week). Output is always a working prototype or a decision memo — never just slides.
Do you sign NDAs and DPAs?
Yes. I work under NDA by default, and I'm comfortable with GDPR / data-processing addenda. I'm based in Poland and invoice as a registered EU entity; VAT details come with the contract.
What if AI isn't actually the right answer?
I'll tell you. Plenty of workflows are better solved with a script, an SQL view, or removing a step entirely. The first deliverable is the honest opportunity map — even if it ends with 'don't ship AI here yet'.
When is forward deployment a better fit?
If you'd rather have an engineer embedded in your team shipping into your own production environment than an outside advisor, look at forward deployment (/consulting/forward-deployment). Advisory often becomes the scoping phase for an embedded engagement.
// LET'S TALK
Ready to put AI to work?
One email and we're on a call within a week. I read every message and reply within one business day.