Senior engineer with 6+ years building and owning production AI systems end-to-end — architecture, system design, evaluation, and delivery — across health tech, insurance, legal tech, and aerospace. Now pursuing a PhD, with research interests in foundation models, continual learning, and neuro-symbolic AI.
Experience
Senior AI Engineer — Independent Contractor
- Own architecture, system design, and tech-stack decisions for production AI — backend, infrastructure, and the evaluation pipelines behind RAG systems and agents.
- Bosch Legends Lab — architecture owner of an AI expert-sourcing platform; introduced Temporal workflow orchestration to replace an event-driven design unfit for long-running, human-in-the-loop AI, and led a microservices → modular-monolith migration. Built CV parsing, profile generation, a web-exploration research agent, and a hybrid-search + re-rank + RAG expert matcher.
- Novee AI — architected LLM structured-extraction systems for insurance underwriting (energy-asset, war/terrorism perils, general risk); built the domain-specific evaluation frameworks, failure-mode loops, and human-in-the-loop labeling that anchor iteration.
- Genie AI — prototyped LLM agents for a legal-writing assistant.
AI Software Engineer — Peppy Health
- Core contributor and senior engineer on the AI platform that became Euphoric, now a standalone company.
- Designed and built RAG systems — a benefits navigator and health chatbot — with FastAPI, OpenAI, pgvector, and Weaviate.
- Owned evaluation as the product’s backbone: pipelines combining classical IR metrics, domain measures, and expert feedback; applied A/B testing and statistical significance to LLM features.
- Ran internal knowledge-sharing on information retrieval (keyword → hybrid search → re-ranking).
Senior Software Engineer — Tractable
- Delivered a production ML system for insurance claim-leakage assessment that directly enabled expansion into new European markets.
Software Development Engineer — Amazon Web Services
- Highly available, low-latency web services for cloud infrastructure.
Machine Learning Engineer — Mediate (MIT spin-off)
- Founding engineer on Supersense, an assistive-tech app for blind and visually impaired users; established the ML and mobile architecture end-to-end.
- Built a deep-learning document edge-detection system that outperformed classical computer vision, a MobileNet intent classifier, and a Core ML 3D document-orientation system; deployed on-device with TensorFlow Lite and Core ML.
R&D Engineer — Turkish Aerospace Industries
- Built in-house flight-test data analysis and validation software (Python, NumPy/SciPy/Pandas, MATLAB), replacing costly tooling and improving team productivity.
Education
PhD, Information Systems — METU · Deep learning, language models, reinforcement learning, information retrieval.
MSc, Data Informatics — METU · Thesis: Knowledge Graph Augmented Multi-Hop Question Answering Using Large Language Models.
BSc, Mechanical Engineering — METU ·
Skills
- AI engineering
- Production AI products end-to-end — RAG systems, agents, and LLM fine-tuning, serving, and post-training.
- Evaluation
- Systematic evaluation of AI systems — evaluation pipelines, offline/online metrics, human/judge alignment, and statistical rigor.
- Architecture
- System design and architecture for AI products; backend, infrastructure, and observability.
- Research
- Rigorous experimental methodology — strong baselines, ablations, and honest measurement.
- Leadership
- Technical direction, tech-stack decisions, and cross-functional collaboration.
Awards
- DocVQA challenge winner — ICDAR 2026.