▸ Led AI feature ideation and engineered orchestration systems across multiple products. Built an LLM-powered assistant for event planning with a RAG pipeline integrated with MongoDB.

▸ Leveraged Azure AI Foundry and LangChain to deploy working GenAI agents, utilizing LangSmith for prompt management, versioning, and AI response evaluation loops.

▸ Built an automated live event direction system: fine-tuned RF-DETR for object detection (mAP ~48 → ~51) and implemented persistent multi-person tracking on LiveKit WebRTC streams.

▸ Designed a multi-agent LLM decision-support prototype (LangChain + custom Python orchestration) with RAG over user-uploaded documents (FAISS) and local-first web search via SearXNG.

▸ Built a "Proactive HMI" multi-objective navigation and POI search prototype (Unity + HERE Maps): translated natural language user intents into weighted-scoring and rule-constrained ranking.

▸ Patent filed (co-inventor): "Proactive Combined Search for Navigation Point-of-Interests (POIs)" — Fraunhofer IAO.

▸ Built an XR prototype for medical education on Microsoft HoloLens (Unity/MRTK), streaming microscope feeds via WebRTC.

▸ Implemented eye-tracking features (gaze cursor, gaze logging, ROI analysis) synchronized with video.

▸ Developed embedded C firmware for ZigBee lighting control.

▸ Integrated proprietary sensor board with ZigBee SoC and validated full system behavior.