// Data Scientist & ML Engineer · Budapest
I build things that think.
Recommender systems, a Slay-the-Spire AI, and the full-stack apps that ship them. I care about models that work in production — fast, measured, and honest about their numbers.
95%
memory reduction (Go vs Python)
0.986 R²
solar forecast model
1200+
tests passing on the STS bot
37k
anime in the recommender
About
Background, the tools I reach for, and how I work.
Work
Recommenders, a game-playing AI, automation pipelines.
Play
Word games I built in the browser. Go on, take a guess.
Featured
Anime Recommendation Engine
A Go backend serving multi-signal recommendations over 28k+ anime in ~58 MB of memory and 60 ms per query — a 95% memory cut over the Python version. Semantic + behavioral embeddings, MMR diversity, and a franchise graph, with a React front end.
Go embeddings FastAPI Redis Docker React
View project
Tools I reach for
Python
Pandas
TensorFlow
SQL
Rust
FastAPI
Docker
Redis
Spark
Kafka
React
Git