We combine modern AI with hands-on knowledge of financial regulation: KNF, EBA, IFRS 9 and the AI Act. We build machine learning models and generative AI solutions for banks, fintechs, lenders and insurers so that they pass validation, audit and supervisory review.
Ready-made packages with a fixed scope, timeline and price. You know from the start what you will receive, when, and what it costs.
For: Lenders, BNPL providers, fintechs and banks
A machine learning scoring model ready for the new consumer credit rules (CCD2) and AI Act requirements.
For: Banks, insurers, lenders
Independent validation of a PD, LGD, ECL, scoring or machine learning model, in line with KNF Recommendation W.
For: Banks, insurers, fintechs
A review of the AI systems in your organisation and a compliance plan ahead of the high-risk deadline of 2 December 2027.
For: SME banking, leasing, factoring
An AI agent reads financial statements and credit documents, extracts the data and feeds a rating model. We cross-check the results against the Aminer database.
For associations of cooperative banks we deliver shared models, validations and training for many banks under a single contract.
We work with management boards, data teams and risk departments of financial institutions. We help choose the AI use cases that pay off, build them and deploy them so that they pass validation, audit and supervisory review.
A review of processes and data, selection of AI use cases with a quantified business impact, a deployment roadmap and AI governance rules. We draw on our experience with the Big Data strategy of a large bank.
AI agents that read financial statements and credit documents, extract the data and feed rating models, cross-checked against the Aminer database. Plus assistants that work on company documents (RAG) and back-office automation, with answer quality evaluation and data protection.
CCD2-ready scoring models for fintechs, BNPL providers and lenders, rating models, and predictive models for sales, pricing, customer churn, forecasting and fraud detection. Built with classical methods and AI, always compared with a simpler benchmark.
Independent validation of statistical, machine learning and language models: methodology, data, stability, interpretability and robustness tests, in line with KNF Recommendation W. Plus monitoring of models in production and data drift detection.
Inventory and classification of AI systems, requirements for high-risk systems, technical documentation, human oversight, fairness testing and explainable decisions.
Workshops for boards, managers and teams: how AI works, where it pays off, what risks it brings and how to control them. Technical training in machine learning and interpretability. They also support the AI literacy obligation under Article 4 of the AI Act.
PD, LGD and EAD models, the IRB approach and CRR3 requirements, IFRS 9 and ECL, stress testing and risk function reviews, increasingly with machine learning.
Benchmarks and industry indices built on data about Polish companies, market potential forecasts, automated collection of data from websites.
An AI deployment succeeds when the model delivers a measurable effect and behaves predictably. That is why we compare every model with a simpler benchmark, explain its decisions and test it the way a validator, an auditor and a supervisor would.
Discuss an AI deploymentExplaining a single prediction
We compare every model with a simpler benchmark and measure its impact on business results, so the gain from AI is measured, not assumed.
Global and single-decision explanations: SHAP, PDP, ICE and ALE. In lending they underpin the explanations owed to customers under Polish banking law and CCD2; in sales and service they show what to do with a prediction.
Obligations depend on a system’s risk. Creditworthiness assessment and credit scoring of natural persons are high-risk systems, with requirements applying from 2 December 2027. We prepare system classification, technical documentation, human oversight and post-market monitoring.
Evaluation of language model answers, tests for hallucinations and prompt injection, data access control and rules for using AI tools in the company.
We check whether a model treats any customer group unfavourably, how it reacts to data drift and whether its explanations stay stable over time. The AI Act requires data sets to be examined for possible bias.
A model changes with the market and the customers. We set up quality and data drift monitoring, alert thresholds and a revalidation plan, so you know when a model needs attention.
A typical machine learning model project takes 8 weeks from data handover. Our AI tool speeds up development and testing, and we write the documentation alongside the model rather than at the end. We meet the client team every week and agree acceptance criteria before development starts.
Selected engagements. We do not name clients without their consent.
Aminer maintains its own database of Polish companies, integrated from public registers. We use it in client analyses and to enrich risk models.
For a bank, we prepared market loss-rate benchmarks for business client segments and assessed the credit risk of a sample of companies with a rating model built on Aminer data.
Automated assessment of Polish companies based on electronic financial statements, alerts on register changes and an explanation of every score. For leasing, factoring and credit control teams.
Ask about early accessWe deliver analysis results in aggregated form, without personal data. How we process register data is described in the notice for data subjects.
Ask about data or an analysisAminer combines AI and data analytics consulting with long experience in finance and its own database of Polish companies. Founded in 2020, it works for banks, insurers, energy companies and the real estate sector.
Founder and CEO
He specialises in machine learning, model interpretability and deploying AI in financial institutions. He has over 25 years of experience in finance: he started in the Risk Management Department of Kredyt Bank (KBC Group) and then spent 14 years at PwC, KPMG and EY, leading risk modelling and data analytics teams, at EY as an Associate Partner.
He led the Big Data strategy of a top-5 bank, pricing and predictive modelling projects and major regulatory implementations in Polish banks. In 2020 he founded Aminer, where he combines AI delivery with the company’s own database of Polish companies.
He has been with the Faculty of Mathematics and Information Science at Warsaw University of Technology since 1994, where he now teaches interpretability and explainability of AI models.
Tell us briefly about your project or question. We will reply to the email address you provide.