DS360 — Full-Stack & ML Engineer
July 2024 – Present
🛠️ Expansion Models — a production ML platform deciding where a retail chain opens its next store. Four models behind one API: predicted turnover of a new location, cannibalisation of the stores around it, twin-store lookup and seasonality.
- ML service — Flask + flask-smorest REST API with an OpenAPI/Swagger contract and marshmallow validation, serving LightGBM and scikit-learn models (regression, classification, similarity search) together with prediction intervals. Containerised and deployed to Cloud Run across dev/stg/prod behind Identity-Aware Proxy, with a one-command deployment script (Cloud Build → Artifact Registry → Cloud Run).
- Network simulation — one request scoring several openings, relocations and closures at once: cannibalisation composed across overlapping catchment areas, turnover of the affected network before and after the change, and twin-store suggestions for every new location.
- Geospatial feature engineering — a bronze/silver/gold pipeline on PostgreSQL + PostGIS and BigQuery: OpenStreetMap data, walk- and drive-time isochrones from an external service, POI density and distance decay — several hundred features per location.
- MLOps — model artefacts versioned in GCS, every prediction historised to BigQuery for monitoring and back-testing, structured logging in Cloud Logging, scheduled data refresh in Cloud Composer (Airflow).
📊 Project Management — led the migration of a client's suite of web applications: independent tools serving different departments under one solution, running on an ageing PHP version and on VMs that kept filling their own disks until they died. They now run as containers on Hetzner against a shared database. The work also covered the environment itself: configuration, and fixes to code and outdated libraries inherited from the old stack. Keeping every application reachable throughout meant designing the cutover itself: how the new environment goes live, and how all of it returns to the previous revision the moment anything misbehaves. Deployment windows were at night, which made coordination the harder half of the job: three people at DS360 on my side, six on the client's, run jointly with their own project manager.
🎮 Simulation Development — extended an airport simulator used for air-traffic-control training (Unity, 3D) with ground vehicle traffic driving across the airport model, giving instructors a new class of events to train on.
🎨 Frontend Development — reworked the structure and code of an existing Vue.js website for a consulting firm and adapted it to new requirements; deployed to Cloud Run through automated CI/CD from GitHub.
Python Flask LightGBM scikit-learn pandas PostgreSQL PostGIS BigQuery Docker Cloud Run Cloud Build Cloud Composer Vue.js Unity CI/CD
- MSc Eng. (mgr inż.) — Computer Science (spec. Intelligent Internet Technologies), Bialystok University of Technology · 2026
- BEng (inż.) — Computer Science, Bialystok University of Technology · 2024
- IT Technician (technik informatyk) — technical secondary school · 2020
- HTML, CSS, Bootstrap, SCSS/SASS, Tailwind
- JavaScript, jQuery, Axios, TypeScript, Angular, Vue, React
- C# (.NET) — LINQ, EF Core, WPF, ASP.NET Core (Web API, MVC, Razor Pages)
- Python — Django, Flask, SQLAlchemy, psycopg
- Data Science & Machine Learning: Scikit-Learn, LightGBM, TensorFlow, Pandas, Jupyter Notebook
- C/C++
- Rust
- PHP
- SQL
- Regex
- Design Patterns, Clean Code
- RESTful APIs (incl. HATEOAS)
- Geospatial Data Processing (Python, PostGIS, BigQuery, OSM)
- PostgreSQL (incl. PostGIS)
- Microsoft SQL Server
- Oracle SQL
- MySQL | MariaDB
- SQLite
- MongoDB
- Google Cloud Platform — Cloud Run, Cloud SQL, Cloud Build, Artifact Registry, GCS, BigQuery, Cloud Composer
- Docker, Virtual Machines (VM)
- CI/CD
- Unity 2D/3D
- Godot
- Unreal Engine 5
- 🪟 Windows, 🐧 Linux, 🍎 macOS
- Visual Studio, VS Code
- Webpack, Vite
- Swagger, Postman
- SSMS, Oracle SQL Developer, DBeaver
- GitHub Desktop, Pages, Codespaces, Actions, Copilot
- Azure DevOps (TFS)
