I'm passionate about applying AI and Data Science to real-world challenges.
As Head of Mentors & Coach at RoboCode, I lead and mentor teams in Robotics, AI, and Computer Vision, fostering innovation through hands-on, project-based learning. I also serve as President of the Industrial Data Science Student Research Group at AGH University of Krakow, where we develop an AI-controlled solar farm, home security system based on SMM and agents and digital twin systems.
My experience blends Machine Learning, Data Analytics, IoT, and Leadership β connecting technology with practical impact.
Core & Backend
AI, LLM & Computer Vision
Robotics & IoT
MLOps, DevOps & Cloud
Data Viz & Dashboards
| π Project | π οΈ Tech Stack & Details |
|---|---|
| Parcel Damage Detection End-to-end Computer Vision system for logistics. π Read Paper (IEEE/ResearchGate) |
YOLOv11 PyTorch ONNX Runtime MLFlow Docker Dash (Plotly) FiftyOne |
| RepoSense MCP Production-ready MCP server that gives AI assistants deep access to GitHub repos β smart file ranking, stack detection, code search, and 14 built-in analysis prompts. Dual runtime: Python (self-hosted) & Cloudflare Worker (edge, Streamable HTTP). π Docs |
MCP TypeScript Cloudflare Workers Durable Objects FastAPI Python OAuth 2.0 SSE Streamable HTTP GitHub API |
| AISpider β Azure MLOps & Predictive Maintenance π (Private / In Development) Quadruped robot with custom gait algorithms, on-board Vision LM, and a full Azure MLOps pipeline for predictive maintenance. Predicts top-5 sensor failures 30 days ahead (servos, LiDAR, sonar, IMU, ToF). ML feature pipelines with tsfresh, SMOTE balancing, and SHAP explainability. Deployed via Azure ML Managed Online Endpoints with canary releases, Azure DevOps CI/CD (Recall >90% gate), Azure Monitor + Data Drift monitoring, and Terraform multi-env infrastructure. |
Nvidia Jetson AzureML Azure DevOps MLflow LightGBM tsfresh SHAP Terraform MicroPython Ollama MQTT Inverse Kinematics FastAPI |
| Kahoot Quiz Architect GenAI tool that automates quiz creation from PDFs. Features strict data validation and structured outputs. |
Google Gemini API Streamlit Pydantic Python GitHub Actions OpenPyXL |
| Hybrid Neuromorphic IoT System π¬ (Research Phase @ IDS AGH) Developing an ultra-low-power event-driven architecture. Concept: Analog SNN neurons (mW range) detect environmental anomalies to trigger a high-level digital LLM agent for action. |
Spiking Neural Networks Analog Hardware LLM Agents Edge AI IoT |
-
π€ Head of Mentors @ RoboCode
- Leading and mentoring teams in Robotics, AI, and Computer Vision.
- Fostering innovation through hands-on, project-based learning methodologies.
- Organizing technical workshops and debugging hardware/software integrations for students.
- Developing hardware/software for quadruped robots (AISpider).
-
π President @ Industrial Data Science Group
- Building "Digital Twins" for industrial sensors.
- Organizing workshops on Python, Cloud, and Data Engineering.
-
π IEEE Author
- First autor on research on Automated Parcel Damage Recognition, blending machine learning with practical industrial impact.
