rizgarozan.github.io · LinkedIn · itch.io · Hugging Face · rizgarozan7@gmail.com
I build games, tools and AI 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. This summer I was a software intern at Barko Elektronik, where I set up a local LLM stack on company hardware, wired it to Active Directory, and built FastAPI services in Docker that feed Oracle data into live Power BI dashboards.
158 pull requests merged into 41 projects I don't own — mteb, kornia, unity-mcp and docling among them — with 17 more in review.
|
|
|
|
|
|
|
|
|
Fixes sent back to the retrieval and Unity tooling I use, refreshed daily by a workflow.
| Pull request | Project | |
|---|---|---|
| 🟣 merged | Accept a string main_score_function in NanoBEIREvaluator | huggingface/sentence-transformers · ★ 19.3k |
| 🟣 merged | fix(custom-tools): keep handler annotations through wrappers on Python 3.14 | CoplayDev/unity-mcp · ★ 15.8k |
| 🟣 merged | fix: make the multilabel classification evaluation cap work | embeddings-benchmark/mteb · ★ 3.5k |
| 🟣 merged | tests: write the citation cache as indented JSON | embeddings-benchmark/mteb · ★ 3.5k |
| 🟣 merged | dataset: add TurkishRAGEvalRetrieval (Turkish passage retrieval, health) | embeddings-benchmark/mteb · ★ 3.5k |
…and 153 more merged pull requests. Merges by project: mteb 23 · kornia 18 · snapclaim 13 · unity-mcp 9 · docling 8 · telegram-mcp 6 · bugbug 6 · module-replacements 6 · results 5 · browser-compat-data 5, and 31 more projects.
- Turkish RAG Eval 0.2.0 is on PyPI: 300 queries (58 hand-labelled, 242 LLM-drafted and double-labelled) across five models, with the dataset on Hugging Face. The dataset is now an MTEB task, TurkishRAGEvalRetrieval (mteb#5615, merged), and the drafted queries are open for human review as good first issues.
- Lockstep Tactics is playable: a C++20 tactics game checked on every push to be bit-identical between MSVC x64, Apple clang arm64 and WebAssembly, with an integer-only AI that wins 998 of 1,000 matches against random orders. Play it in the browser against the AI or against a friend in another browser: the two run in lockstep over WebRTC and compare hashes as they go.
- Graduating in June 2027; open to internships and new-grad roles in applied AI and game engineering.




