I build and run a self-hosted, air-gapped LLM platform at Roketsan that 200 quality engineers use daily for FMEAs, inspection plans, V&V traceability matrices, document control, and code compliance checks.
I designed the platform end to end around open-weight models served with vLLM, using quantization, batching, scheduling, schema validation, access control, and usage tracking to support multiple applications on one on-prem GPU server.
My work combines GenAI with applied ML for aerospace hardware: weld-defect and rocket-motor void detection from X-ray imagery, anomaly detection on rocket-motor test data, and supplier delivery and quality-risk prediction. I also led avionics design quality assurance for Türkiye’s Satellite Launch Vehicle Program, which directly shaped the quality-engineering tools I build today.
Outside proprietary aerospace work, I publish GenAI, RAG, and ML projects on GitHub and earned Kaggle Notebooks Master tier with one gold and two bronze competition medals. I also built fmeatool.ai, an agentic LLM application that generates AIAG & VDA-compliant FMEAs from technical documents with Excel export.
