rizgarozan.github.io · LinkedIn · itch.io · Hugging Face · rizgarozan7@gmail.com

I build measurement systems: retrieval evaluations that say which part of a RAG pipeline earns its cost, and game tools that score a level or catch a broken font before anyone ships it. Final-year Computer Education & Instructional Technology student at Hacettepe University, Ankara.

63 pull requests merged into 19 projects I don't own — mteb, kornia, LightRAG and unity-mcp among them — with 31 more in review.

Five projects, in the order I'd show them to you. Every number on a card is copied from the project's own committed results.

_{Turkish RAG Eval: 58 hand-labelled queries, published as a dataset on Hugging Face; differences under 0.05 nDCG are noise, and the README says so. Match3 Lab plays in the browser and ships 50 tests and a decision record for every rule. TMP Glyph Audit: 24 xUnit + 5 EditMode tests, on OpenUPM.}

Fixes sent back to the retrieval and Unity tooling I use. The AI and Unity projects get a row each, the rest are counted; a workflow refreshes this every week.

_{…and 24 more merged pull requests to Effect-TS, ManimCommunity, SchemaStore, TheAlgorithms, e18e, embeddings-benchmark, mdn, modelscope, pylint-dev, radzenhq, umbraco.}

• Growing Turkish RAG Eval from 58 hand-labelled queries to a 300-query, six-model benchmark, with the dataset on Hugging Face.
• A few pull requests a day to the retrieval and Unity tooling in the table above.
• Graduating in June 2027; open to internships and new-grad roles in applied AI and game engineering.

_{Header and cards are SVGs rendered by scripts/ in the site's palette; the cards' text is outlined so they look the same on every machine.}