software engineer at mobiliz. rag pipelines, vector search, and self-hosted llms in production — shipped, not demoed.
i'm a software engineer who likes systems that reason — retrieval-augmented pipelines, vector stores, and the small unglamorous infrastructure that holds them together. i work in python, typescript, and java. i prefer writing code that i'll still understand in six months.
at mobiliz i'm building ai features for fleet telematics and enterprise document search: self-hosted llms, retrieval that has to stay honest, websocket streams, docker everything. before that, backends at siemens. the through-line is data — how it moves, where it rests, what you ask of it.
off the clock i read finance and psychology (lately: thinking in bets and psychology of money), i lift heavy-ish, run slow-ish, and i play pickup basketball whenever ankara isn't too cold.
building ai features at fleet-telematics scale: rag over turkish enterprise documents, vector store migrated faiss → qdrant, generation moved onto self-hosted qwen3.6 35b-a3b behind vllm. multi-container docker: fastapi · react · mongodb · clickhouse. embedded superset dashboards with row-level security. websocket real-time, jwt auth, leaflet dashboards.
multi-threaded and async pipelines for concurrent report generation. load balancing for workload distribution, dynamic routing via AOP, improved error handling and logging.
internal apps for power-system analysis and reporting. full-stack with c# / .net core / angular / typescript.
if you're building something interesting — especially if it touches ai, retrieval, or technical infrastructure — i'd like to hear about it.