[ 00 ] — AI Engineering · Data Architecture
We take GenAI pilots to production: agentic systems and RAG, LLM observability, MLOps. Behind the delivery stands well over a decade in Data Science / ML — we work reproducibly and auditably, and the IP stays on your side.
[ 01 ] — SERVICES
A named scope, a fixed timeframe, a concrete artifact at the end. We give a fixed quote after a 30-minute scoping call.
For companies that want proof of value before investing in a platform.
We define the business metric right at the start — so your project doesn’t join the vast majority of GenAI pilots that show no measurable return (MIT). If the results don’t defend the metric, you get an honest report: why, and what’s next.
Get in touch →For teams running LLMs in production with no visibility into quality and cost.
An audit trail and evaluation records for EU AI Act requirements. Transparency obligations (Art. 50) have applied since 08.2026; requirements for high-risk systems were moved by the digital omnibus to 12.2027 (Annex III) and 08.2028 (Annex I). The technical groundwork takes longer to build than that runway suggests. We deliver it — we leave interpreting the regulations to lawyers.
Get in touch →For boards and CTOs ahead of an investment decision.
An independent audit: you get the document, the backlog and the estimate — you can build with anyone. We don’t upsell ourselves an implementation “on the side”.
Get in touch →For DS teams with notebooks that need to deliver repeatably.
More than half of AI pilots never reach production (Gartner). We close that gap: from notebook to a reproducible, auditable pipeline.
Get in touch →For companies whose supplier, product, customer and technology dictionaries have drifted apart across systems.
“Cleaned up” is not a claim — it’s a validator’s output: a report of merged and rejected pairs with rationales, plus structural tests in CI that keep guarding quality after we leave.
Get in touch →For companies that need a senior GenAI engineer on an ongoing, flexible basis.
Max. 2–3 clients in parallel — exclusivity instead of agency scale. Piotr builds himself, not just advises.
Get in touch →[ 02 ] — FOR WHOM
You have a GenAI PoC that hallucinates or doesn’t deliver ROI. We take it to production — or tell you plainly that it isn’t worth pursuing, and why.
You added AI to your offering, but you lack a senior GenAI engineer at production level. We bring the AI layer under your brand — project subcontracting, not staff leasing.
HR-tech, scoring, fintech — you need the technical groundwork for compliance: logging, documentation, evaluation and an audit trail.
[ 03 ] — PROCESS
A free consultation. We define the problem and the success criteria.
Architecture, milestones, estimate. You know what you’re paying for.
Every stage ends in a working product. You accept it — you pay.
The IP, documentation and reproducible pipelines stay with you.
[ 04 ] — STACK
Python · SQL · pandas / NumPy · scikit-learn / XGBoost · FastAPI · REST / API · Docker · Kubernetes · Git · CI/CD · Airflow · MLflow · Azure · GCP · LangChain
[ 05 ] — WORK
Independent Datarmination work — our own methods and tools. Every number below can be recomputed from project artifacts.
Problem
No coherent, searchable competency structure from job postings.
Approach
A multi-level taxonomy covering thousands of competencies, built via multi-model consensus: different models independently review every disputed assignment, with the outcome decided by majority vote — and every decision recorded.
Value
An auditable hierarchy with a full voting trail and a cross-domain flow map. Low model agreement on disputed cases is proof that a single model is not enough.
Problem
Fragmented, inconsistent AI terminology with no links between concepts.
Approach
Entries gathered from many independent model sources across successive rounds and linked by a graph of typed relations. In the final round a substantial share of candidates failed a two-stage source verification.
Value
A reproducible reference glossary: source links after automated validation, a maturity rating for every entry, concept evolution chains — fully regenerable from code.
Problem
Agentic systems that guess instead of admitting they don’t know.
Approach
Several specialized agents (embedding-based routing, multi-competency splitting, synonyms, taxonomy, translations) + deterministic integrity validation without an LLM. The same system built twice — on OpenAI Agents SDK and LangGraph — with a framework comparison before committing.
Value
Orchestration in code, not by an LLM: closed action codes, confidence thresholds, a human-review path and dozens of tests with a mocked model — an architecture built for production, not for a demo.
Problem
A single research model hallucinates and misses sources.
Approach
The same research run in parallel across several independent deep-research systems; a fact counts as certain only once confirmed across independent tracks. Anti-hallucination regime: hard rules, including a mandatory source URL for every claim.
Value
A reusable research template: reports across several thematic tracks, cited sources, documented divergences between tracks.
[ 06 ] — FIRST CALL
[ 07 ] — FAQ
Because a fair quote depends on your data and scale. We give a fixed price after a 30-minute scoping — upfront, not “as we go”.
The scope of each stage is clear and finite. I build the core myself; for overload I have a consortium model and trusted partners.
Yours. The code, prompts and pipelines in full — we deliver them reproducibly, and you run them yourself.
On acceptance of each milestone. You don’t pay for a stage you haven’t accepted.
NDA as standard, before any conversation about data. We work on your infrastructure wherever data can’t leave it.
Either hardening to production (PoC-to-Production), or an ongoing engagement on a retainer model.
A test set + LLM-as-a-Judge + metrics defined upfront. Without that, “it works” is an opinion, not a fact.
[ 08 ] — CONTACT
Let’s talk about your project — from feasibility analysis to a working demonstrator.
We reply within 24h on business days.