reyhan bayraktar
GenAI Engineer · LLM inference, evaluation and agents
I serve large language models on hardware I can touch, and I measure them until the numbers hold up. Many of my favourite results are speedups that turned out not to exist.
At NewMind AI in Istanbul I built the evaluation system for a production agent platform and brought up around twenty open-weight model families on H200, GB300 and DGX Spark.
selected work
mecellem-legal-bench
An open benchmark that asks whether models apply the law or just cite it. 42 Turkish legal practice questions, 367 rubric criteria, an LLM judge, and a public leaderboard of 16 frontier models. dataset · leaderboard
Benchmarks that lie
A reported 4–5× speculative-decoding speedup turned out to be a comparison against a baseline with CUDA graphs switched off. Same story with a model I had written off as slow. I now check the flags before I believe the number.
Grading agents, not answers
Trajectory-level evaluation for a production agent platform on MCP tools. Each question runs k times, each run is traced end to end, and each step is graded: which tool, in what order, whether the citations are real, and what it cost. It became the release gate for new tool servers.
also
- Served ~20 open-weight model families (Qwen, DeepSeek, Kimi, GLM, MiniMax, Nemotron, Gemma) with vLLM and SGLang in FP8 and NVFP4.
- Multi-node MoE serving on the MareNostrum 5 supercomputer, up to 1T parameters.
- Compared EAGLE-3, MTP, DFlash and DSpark speculative decoding. The gain shrinks with load on dense models and grows on MoE.
- Pretrained a 1.7B Turkish diffusion language model from scratch on MareNostrum 5.
- Multi-agent RAG for legal and financial documents: an orchestrator, up to 16 parallel citation-grounded sub-agents and a reviewer.
publications
- RDP LoRA: Geometry-Driven Identification for Parameter-Efficient Adaptation in Large Language Models
- TurkColBERT: A Benchmark of Dense and Late-Interaction Models for Turkish Information Retrieval
- Turk-LettuceDetect: A Hallucination Detection Model for Turkish RAG Applications
- Guided Decoding and Its Critical Role in Retrieval-Augmented Generation
- Automatic Mine Detection for Underwater and Surface Vehicles
experience
- Research Engineer, NewMind AI
- ML Engineer Intern, Voiser
- Image Processing Lead, Auvtech university robotics team
- B.Sc. Computer Engineering, Istanbul Sabahattin Zaim University