Hi! I’m Mikołaj, a Software Engineer II and a student at Warsaw University of Technology, with a passion for music. My professional journey includes an internship at Allegro and Bridgestone and a current position as a Software Engineer II at Box. I’ve worked on large-scale real-time collaborative apps, streaming big data systems, and backend-for-frontend microservices. My technical expertise encompasses Java, DevOps, Kubernetes, Kafka, Spring, and React. Alongside my engineering roles, I’ve competed in coding competitions, securing 1st and 3rd place in the Best Hacking League and 2nd place in the Ensemble AI Hackathon. Academically, I'm focusing on machine learning for music and audio explainability and generative music models.
Thanks for stopping by!
Proficient in Java, Spring Framework, and Node.js for robust application development especially for large-scale real-time collaborative content creation products. Product-minded in making useful software. Knowledgeable in design patterns, TDD, DDD, and microservices for scalable solutions.
Skilled in DevOps with Kubernetes for application deployment and architecture, ensuring scalability. Proficient with GCP systems like Pub/Sub, Dataflow, and Bigtable for data pipelines, with some experience in Kafka for streams.
Active in machine learning research and data science projects focused on music and audio, applying PyTorch in university work. Solid mathematical background with strong algorithms focus. Aiming to utilize statistical models for data-driven insights.
MusicGen Lens enables a totally new way of controlling generative music models beyond prompts by collecting activations with nnsight and applying activation-level interventions (DiffMean, RFM, SAE) for interpretability and steering.
Microservices parcel tracking system (Spring/Java) with carrier integrations, tracking history in database, search, periodic updates, and event-driven services.
Java Spring + JavaFX messaging app with channels/DMs, real-time chat via RabbitMQ, Elasticsearch full-text search, JWT auth, and PostgreSQL backend.
Real-time survey app for events with QR-code access and live dashboards, built in React/TypeScript with Supabase (PostgreSQL).
Python P2P torrent system with coordinator service, client/file server, chunked transfers, and Dockerized networking.
Python console spreadsheet (Excel-like) with cell addressing, formulas over ranges, live recalculation, and CSV load/save.
C++ implementations of classic software design patterns focused on clean architecture, extensibility, and maintainable backend code.
DNA variant annotation platform for splicing analysis with REST API core, worker services for SNV/INDEL processing, and optional UI.
Thermal-guided rocket project using Raspberry Pi, infrared camera input, servo control, and real-time heatmap visualization and tracking.
C++ OpenGL visualization of the Lorenz attractor with real-time animation of chaotic trajectories and 3D rendering.
Unity AR multiplayer game for Android/iOS using Cloud Anchors and Firebase for shared anchors and real-time sync.
Research on real-time neural style transfer for games using Unity integration, G-buffer guidance, distillation, and quantization for fast inference.
PyTorch diffusion model with UNet architecture and Lightning training for image generation on custom datasets (traffic images).
Sparse autoencoders for unsupervised music representation learning on MusicGen/RAVE/YuE activations, producing interpretable audio embeddings.
MusicGen Lens enables a totally new way of controlling generative music models beyond prompts by collecting activations with nnsight and applying activation-level interventions (DiffMean, RFM, SAE) for interpretability and steering.
Interpreter for a custom music programming language with lexer/parser/AST, static typing, and MIDI import/export.
C++ DJ console built with JUCE for real-time audio playback, multi-track mixing, effects (EQ/reverb/delay), tempo sync, and playlist control.
Python JUCE bindings (via pybind11) for building VST3/AU audio plugins, enabling rapid DSP prototyping with C++ build tooling.
Thesis on steering MusicGen with textual inversion for controllable music generation and interpretability.
Hackathon project “Magazyn 4.0” (3rd place, BEST Hacking League): warehouse optimization platform with FastAPI websockets, React Native barcode app, React dashboard, and MongoDB analytics.
Pocket Travel Agent hackathon app with ElevenLabs TTS, multi-language voice UX, AI trip planning (Perplexity + Mistral), and integrations with Google Maps and OpenMeteo via Make.com workflows.
Ensemble 2024 hackathon solution for ML security tasks (Sybil attack + model stealing) with Python notebooks, dataset utilities, and submission pipelines.
Ensemble AI 2025 hackathon codebase with PyTorch pipelines for image embedding regression (task2) and adversarially robust classification (task3), plus notebooks and evaluation scripts.