I'm a developer from Vilnius, Lithuania. I studied programming at SMK College and started out building front-ends with React and TypeScript. Over the last few months most of my time has gone into machine learning.
- Solving ML challenges on Shipd's Project Eris. Every problem comes with a hidden test set and an AI agent baseline to beat, and most of my models are written in PyTorch.
- One challenge I designed was accepted on the platform.
- Small ML experiments of my own, mostly about Lithuanian and small language models. They are the cards above.
What I care about in ML
- Validation that matches how the model will really be tested. A good score on the wrong split means nothing.
- A simple baseline first, then a neural model that has to beat it.
- Code anyone can rerun: fixed seeds, one command, the same result.
Python, PyTorch, NumPy, pandas, scikit-learn, LightGBM, Hugging Face Transformers, TypeScript, React, Next.js