01
Build around a decision.
Start with a repeated, expensive and consequential decision—not with a fashionable model. The product specification is the decision, its context, its acceptable failure modes and the action it should change.
Founder / ML product / Research
I build companies around hard ML problems.
Research depth. Product discipline. Real-world deployment.
I turn algorithms and clinical insight into machine-learning products that survive contact with workflows, regulation and the market. Co-founder & CTO of MIM Fertility. Computer scientist and assistant professor at the University of Warsaw.
01 / EntrepreneurshipCompany building, financing, product thesis and market focus.
02 / ML product deliveryData, validation, workflow integration, regulation and deployment.
03 / Research depthApproximation algorithms, search, explainability and clinical ML.
2015
Building ML companies since
>$30M
R&D funding directed across programs
14+
Major R&D programs led
12
Peer-reviewed publications
~198
Research citations
01 / Operating thesis
Deep tech is a translation business.
A model in a notebook is not a product. A publication is not a company. The value appears in the translation.
The hard work lives between disciplines: choosing the right decision to improve, designing evidence, building the workflow, earning clinical trust and making the economics work. I operate across those boundaries because that is where most ML products either become real—or die.
02
Use research for leverage.
Deep research matters when it creates a non-obvious advantage: better reliability, stronger evidence, lower cost, more explainability or a product others cannot easily reproduce. It is an engine, not a credibility layer.
03
Deploy to compound.
Real-world use creates the feedback loop. Deployment produces new data, exposes edge cases, sharpens the research question and makes the next product iteration materially better.
02 / How I build
From research question to operating product.
ML products are systems. The model is one component in a chain of evidence, software, workflow and trust.
My work spans the full chain: problem framing, technical architecture, research, funded R&D, product development, clinical validation and deployment. That breadth is deliberate—the hand-offs are usually the risk.
01
Decision map
Find the high-value decision, who makes it, which inputs are available and what a better answer changes operationally.
Output / measurable product specification
02
Evidence design
Define the dataset, baseline, target metric, failure conditions and validation protocol before optimizing the model.
Output / data strategy + proof standard
03
Research edge
Use algorithms, modeling and domain knowledge to create a defensible improvement—not merely a benchmark gain.
Output / non-obvious technical advantage
04
Product system
Build interfaces, workflow integration, monitoring, exception handling and human oversight around the model.
Output / usable, observable system
05
Validation & regulation
Prove performance in the environment that matters and design the evidence package for clinical, regulatory and commercial trust.
Output / decision-grade evidence
06
Deployment loop
Ship, measure adoption, capture edge cases and feed field learning back into both the product and the research roadmap.
Output / compounding product advantage
Most teams stop at 03. The company is built in 04–06.
03 / Ventures
Two companies. One thesis.
Company 01 / Vertical deep tech
MIM Fertility
Artificial intelligence across the IVF pathway—follicle monitoring, endometrial assessment, embryo selection and sperm analysis. Built with fertility clinics to standardize and automate high-stakes clinical workflows.
FOLLISCAN
FOLLISCAN HOME
EMBRYOAID / CE-marked
ENDOSCAN
CASA 3D
AI4IVF
EmbryoGeneScan
Company 02 / Applied ML platform
MIM Solutions
The first spin-off of the University of Warsaw Algorithms Group. A multidisciplinary ML company delivering bespoke deep-learning systems in medicine, e-commerce, security and automotive—with the research culture and engineering depth to tackle non-standard problems.
Healthcare AI
Recommenders
Computer vision
Explainable ML
Predictive systems
04 / Research
Research is the engine, not the garnish.
The theoretical work and the product work are not separate careers. They inform each other.
The research path runs from approximation algorithms and nearest-neighbor search to explainable ML and reproductive medicine. The common thread is reliable decision support under real constraints.
Optimization & algorithms
Approximation algorithms, LP and iterative rounding, parameterized algorithms and practical implementations through PAAL.
Search & explanation
Locality-sensitive hashing and approximate nearest neighbors without false negatives; Shapley and Banzhaf feature importance for tree models.
Clinical machine learning
Personalized prediction and imaging systems in reproductive medicine, where performance must translate into clinical decisions and workflows.
2026
Evaluating the concordance between AI-based and conventional embryo selection: implications for clinical decision-making
How AI-based embryo selection aligns with conventional assessment, and what the gaps mean for clinical decisions.
Reproductive BioMedicine Online
2026
An artificial intelligence platform for automated measurement and count estimation of ovarian follicles during ovarian stimulation and IVF: a multicenter study
Multicenter validation of automated follicle measurement and counting during ovarian stimulation.
J. Assisted Reproduction & Genetics
2023
Personalized prediction of the secondary oocytes number after ovarian stimulation
Clinical and genetic data used to personalize a high-value prediction in reproductive medicine.
PLoS Computational Biology
2022
Improved feature-importance computation for tree models based on the Banzhaf value
More efficient feature-importance computation for widely used tree-based models.
UAI 2022
2020
The PACE 2020 Parameterized Algorithms & Computational Experiments Challenge: Treedepth
Algorithm engineering and empirical evaluation around a hard graph problem.
IPEC 2020
2017
Approximate nearest neighbors search without false negatives for ℓ₂
Similarity search designed around a strict reliability requirement: no false negatives.
ISAAC 2017
2017
Why do cascade sizes follow a power law?
A network-science view of how information cascades grow and distribute.
WWW 2017
2016
Locality-sensitive hashing without false negatives for ℓₚ
Extending reliable similarity search beyond the standard Euclidean setting.
COCOON 2016
05 / Projects
A record of funded, shipped and led work.
Company building in deep tech is a portfolio of hard bets, not one model.
Selected programs across medical AI, predictive systems, security, e-commerce and algorithm engineering. Roles span founder, CEO, CTO, Head of R&D, project leader, principal investigator and lead developer.
| Project | Role | Funder | Years | Funding ≈USD |
|---|---|---|---|---|
| FOLLISCAN HOMEAI support for self-administered transvaginal ultrasound. | Head of R&D & CTO | FENG / PARP | 2024–29 | $5.1M |
| ASFiTET — AcoraiNon-invasive intracardiac-pressure monitoring for heart-failure patients. | Co-founder & CEO | Eurostars 3 | 2024 | $0.5M |
| CASA 3D3D microscope for AI-driven sperm analysis and selection. | Co-founder & CEO | Eurostars 3 | 2023–26 | $0.8M |
| ENDOSCANAI support for ultrasound diagnosis of endometrial receptivity in IVF. | Co-founder & CEO | FENG / PARP | 2023–27 | $3.2M |
| FOLLISCANAI support for ultrasound diagnostics of ovarian reserve. | Head of R&D & CEO | NCBR / POIR | 2021–23 | $2.0M |
| EMBRYOAIDEmbryo viability assessment from image data. | Head of R&D & CEO | NCBR / POIR | 2021–23 | $1.9M |
| User segmentation platformPersonality-based ad segmentation. | Project Leader & CEO | POIR | 2020 | $2.2M |
| ANN without false negativesNearest-neighbor search with no false negatives. | Principal Investigator | NCN | 2019–21 | $20K |
| Flight scoring & predictionAdaptive ML for flight delay and compensation forecasting. | Project Leader & CEO | POIR | 2018–20 | $3.8M |
| Ad categorization systemAutomatic categorization into more than 20,000 categories. | Project Leader & CEO | Gemius | 2017–18 | — |
| PROKRYMCrime-prediction system for the Polish police. | ML Team Leader | NCBR | 2015–18 | $1.8M |
| PAAL Proof of ConceptGame-theoretic user-behaviour models for e-markets. | Programmers' Lead | ERC / #680912 | 2015–17 | $0.7M |
| RTB CTR predictionClick-through-rate prediction for real-time bidding. | Project Leader | RTB House | 2015–16 | +14% revenue |
| PAALPractical Approximation Algorithms Library, open-source C++14. | Lead Developer | ERC / #259515 | 2013–15 | $4.7M |
06 / Speaking & recognition
Speaking and recognition. Useful context, not the story.
Selected talks
Multi-Center Validation of an Integrated AI Platform Across Three Sequential IVF Decision Points
ASRM Scientific Congress · Baltimore, USA
Non-Invasive AI-Based Euploidy Prediction from Day-5 Blastocyst Images
ASRM Scientific Congress · Baltimore, USA
Categorical Decision Concordance Between an AI Endometrial-Assessment Platform and Expert Clinicians
MRSi · Chicago, USA · oral abstract winner
AI-Driven Patient Intake: Fewer Questions, Better Visits
Health AI & CyberSec Summit · Warsaw
Keynote — “Are we really using AI?”
27th CRB Symposium · AAB College of Reproductive Biology · Las Vegas, USA
Validation of AI in Follicles
TSRM — Turkish Society of Reproductive Medicine · Antalya, Turkey
Prospective Validation of an AI Platform for Automated Follicle Counting
ESHRE 40th Annual Meeting · Amsterdam · oral abstract O-218 (Human Reproduction)
FOLLISCAN: Automating Ultrasound Examination with Deep Learning
Predictive Analytics World · Healthcare · Las Vegas
AI for Fertility Disorders
LSI Europe · Emerging Medtech
Designing C++ Implementations of Complex Combinatorial Algorithms
C++Now · Aspen, Colorado
Recognition
2023
23 Most Influential Poles in AI
My Company Polska
2023
Best Innovation in Fertility Care
Company award · MIM Fertility · Fertility Care Awards, European Fertility Society — FOLLISCAN
2023
Deloitte Technology Fast 50 Central Europe
Company recognition · MIM Fertility
2022
SoDA Awards — Innovation category
Company award · MIM Fertility · FOLLISCAN & EMBRYOAID
2022
Eagle of Innovation
Company award · MIM Solutions · Rzeczpospolita
2022
InCredibles — programme laureate
Company award · MIM Solutions · Sebastian Kulczyk
2022
Rookie of the Year
Company award · MIM Solutions · My Company Polska & AWS
2022
CEBioForum Award — BioDigital
Company award · MIM Solutions
2019
KaggleDays Warsaw — Winner
Offline data-science competition
2016
RecSys Challenge — 2nd Place
ACM RecSys
Service & affiliations
Advisory
Global Gateway Programme
Business Advisory Group to the European Commission President
Faculty
Assistant Professor
University of Warsaw · Maths, Informatics & Mechanics
Member
AIFS
AI Fertility Society
Reviewer
Journal peer review
Human Reproduction · Fertility & Sterility · Systems Biology in Reproductive Medicine · Discover AI
07 / Contact
Bring me a hard decision worth improving.
I am most useful where deep research must become a product: fertility clinics, medical AI, ML infrastructure, deep-tech ventures and research collaborations.
piotrwygockicom