Filip Noworolnik AI researcher • engineer • entrepreneur
From research to real-world AI products

Building practical AI for sports, medicine, and industry.

I’m Filip Noworolnik — a PhD researcher at AGH University in Kraków, Poland, and an AI engineer working at the intersection of computer vision, multimodal AI, and deep learning with PyTorch. My work spans sports analytics, medical imaging, trustworthy AI, and product-focused R&D.

I care about systems that do more than look good in a paper: they should solve real problems, fit operational workflows, and stand a chance of becoming products, partnerships, or startups.

See selected projects Discuss collaboration
About

A researcher who likes to build.

My profile sits between academic R&D, engineering delivery, and venture creation. I enjoy converting ambitious ideas into structured roadmaps, prototypes, experiments, and credible commercialization stories.

Academic depth

PhD research at AGH University of Science and Technology in Kraków, focused on computer vision, multimodal reasoning, and model interpretability. Published at venues like ICCV. My research background helps me assess what is novel, what is reproducible, and what is actually worth pursuing.

Engineering mindset

Comfortable with the full path from data and experimentation to production — Python, PyTorch, Docker, cloud deployment. I value systems that are robust, practical, and measurable.

Commercial instinct

I naturally think in terms of user pain points, product differentiation, proof of value, TRL progression, and go-to-market logic. I like technology that can leave the lab and survive contact with reality.

Technical expertise

Tools and technologies I work with.

Core ML & deep learning

PyTorch TensorFlow scikit-learn Hugging Face OpenCV ONNX CUDA

Languages & infrastructure

Python C++ SQL Bash Docker Git Linux MLflow W&B

Computer vision

Semantic segmentation Object detection Pose estimation Video understanding 3D imaging (CBCT) Vision Transformers

Domains

Medical imaging Sports analytics Explainable AI Foundation models Multimodal AI MLOps
Focus areas

Where I do my best work.

These are the domains and problem types where I can contribute the most - either as a researcher, technical lead, consultant, or builder.

Sports analytics & movement intelligence

AI for football and broader sports settings, especially where lightweight capture setups, actionable feedback, and clear user value matter more than lab-perfect conditions. I am particularly interested in turning video into coaching insight.

Shot technique analysis and feedback systems
Single-camera perception and performance assessment
Grassroots-friendly AI products for clubs and academies

Medical AI & healthcare imaging

I work on image-based systems where quality, interpretability, clinical relevance, and deployment realism matter. The strongest themes in my work include dermatology, echocardiography, confocal microscopy, and dental imaging.

Skin lesion and cell analysis
Echocardiography and foundation-model-style workflows
Tooth and root canal segmentation with product potential

Trustworthy & explainable AI

I am interested in systems that help people understand model behavior, spot failure modes, and use AI in a way that earns trust. This includes concept-based reasoning, anomaly awareness, and interpretable representations.

AI commercialization & R&D strategy

Beyond models, I enjoy shaping the path from research result to market-facing opportunity - clarifying the product story, identifying early adopters, defining deliverables, and making the technical direction legible to partners and investors.

Selected projects

A portfolio built around real use cases.

A representative sample of the kinds of systems, products, and research directions I have been developing.

Sports AI Computer Vision Multimodal Feedback

Single-camera football shot analysis

A system for automatically evaluating shooting technique from regular video and producing actionable feedback. Combines pose estimation, vision-language models, and motion feature analysis. Published at ICCV 2025 Workshop.

Medical AI Dermatology Spin-out Potential

Dermatology and skin imaging solutions

Skin lesion analysis and microscopy-driven cell understanding using segmentation networks and Vision Transformers. The broader goal is to translate strong technical work into clinically meaningful and commercially viable tools.

Cardiac Imaging Foundation Models Healthcare R&D

Echocardiography AI workflows

Scalable AI workflows for echo video understanding, classification, and reporting support. Building foundation-model-inspired approaches with PyTorch that improve generalization and downstream utility in clinical settings.

Dental Imaging 3D SaaS Direction

Dental CBCT segmentation platform

AI-assisted 3D segmentation of teeth and root canals from CBCT scans, with a product vision centered on dental workflows, interactive visualization, and SaaS delivery. A strong example of research meeting vertical software opportunity.

Research & trajectory

From interpretable vision to deployable AI.

My trajectory is shaped by a recurring question: how do we build AI systems that are both technically strong and genuinely useful in the world?

Interpretable computer vision and trustworthy AI

AGH University of Science and Technology, Kraków. Research on concept- and structure-aware methods, spatial reasoning between higher-level visual concepts, and AI systems that do not behave like black boxes by default. PyTorch, Vision Transformers, explainability frameworks.

Medical imaging and domain-specific AI products

Translation of advanced deep learning methods into practical healthcare applications — echocardiography foundation models, dermatology segmentation, dental CBCT analysis. Emphasis on deployment realism, clinical validation, and building AI systems that earn trust.

Commercially oriented AI systems for sports, medicine, and industry

Building toward solutions that can become partnerships, demos, deployments, and startups — not only publications. Especially interested in applied multimodal systems, foundation models for healthcare, and human-centered decision support.

What I optimize for: technically ambitious work with a believable path to adoption.

Selected publications

Assessing the Quality of Soccer Shots from Single-Camera Video with Vision-Language Models and Motion Features

F. Noworolnik, J. Jaworek-Korjakowska. IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2025.

Paper

Diabetic Foot Ulcer Unsupervised Segmentation with Vision Transformers Attention

F. Noworolnik, A. Brodzicki, D. Kucharski, B. Moniak, A. Kostuch, A. Wojcicka, J. Jaworek-Korjakowska. Diabetic Foot Ulcers Grand Challenge (DFUC), MICCAI Workshop, Springer LNCS.

Paper

DFU-Ens: End-to-End Diabetic Foot Ulcer Segmentation Framework with Vision Transformer Based Detection

D. Kucharski, A. Kostuch, F. Noworolnik, A. Brodzicki, J. Jaworek-Korjakowska. Diabetic Foot Ulcers Grand Challenge (DFUC), MICCAI Workshop, Springer LNCS.

Paper 3rd place, DFU Challenge

Full list on Google Scholar.

Work with me

How I can contribute.

I am most useful in settings where ambitious AI ideas need both technical depth and product realism.

Research collaboration

For universities, labs, hospitals, and R&D teams looking for support in computer vision, multimodal AI, explainability, or experimental design.

AI product strategy

For teams shaping an MVP, demo, grant proposal, spin-out path, or commercialization concept around AI-heavy technology.

Technical advisory

For companies that need an experienced perspective on feasibility, model strategy, dataset planning, evaluation, or roadmap prioritization.

Open to select research collaborations, advisory roles, and leadership positions in AI. Based in Kraków, available internationally.