AI Research Engineer · NewMind AI — Istanbul, Türkiye
Large Language Models · Turkish & Legal-Domain NLP · Inference · Retrieval · Agents
I'm an AI Research Engineer at NewMind AI, working across the full LLM lifecycle — pretraining & fine-tuning, high-performance multi-node inference, retrieval / RAG, evaluation, and agentic systems — with a research focus on Turkish and legal-domain NLP.
- Pretrained Turkish language models on the BSC MareNostrum 5 supercomputer (6,000+ GPU-hours)
- Brought up & benchmarked ~20 frontier model families across vLLM / SGLang and FP8 / NVFP4 / GGUF quantization
- Built retrieval / RAG (semantic caching, late-interaction retrieval, GraphRAG) and multi-agent systems
- Designed LLM evaluation frameworks (calibration, LLM-as-judge, adversarial benchmarks)
- RDP LoRA: Geometry-Driven Identification for Parameter-Efficient Adaptation in LLMs — 2026 · arXiv:2604.19321
- TurkColBERT: A Benchmark of Dense and Late-Interaction Models for Turkish Information Retrieval — ACLing 2025, Elsevier Procedia Computer Science · arXiv:2511.16528
- Turk-LettuceDetect: A Hallucination Detection Model for Turkish RAG Applications — FLLM 2025 (IEEE) · arXiv:2509.17671
- Guided Decoding and Its Critical Role in Retrieval-Augmented Generation — SIU 2025 (IEEE) · arXiv:2509.06631
- Automatic Mine Detection for Underwater and Surface Vehicles — ICCIDA 2024, Springer · DOI
