Ahmed Bilal

Designed LangGraph multi-agent workflows and extended MCP for enterprise knowledge retrieval, while modernising the Electron app UI with ShadCN.

• Designed and shipped LangGraph-based multi-agent workflows extracting structured insights from meetings, chats, PDFs, and Notion knowledge bases; consolidated 4+ data sources into a single query interface.
• Architected agents capable of reasoning over both hot channels (real-time sessions) and cold channels (static knowledge bases), delivering context-aware responses with sub-second latency.
• Integrated and extended MCP (Model Context Protocol), enabling dynamic context sharing between agents and live sessions; reduced redundant API calls by eliminating repeated context fetches.
• Refactored the Electron desktop app, replacing legacy UI with ShadCN; reduced component inconsistencies and cut new feature UI development time.

Built and launched a live B2C AI dental diagnostic platform serving patient X-ray tracking and dental student training verticals.

• Engineered an end-to-end YOLOv8 pipeline for dental X-ray analysis, detecting 4+ dental conditions with >90% sensitivity, reducing false negatives by 31% vs. unaided dentists.
• Built a patient-facing portal enabling X-ray upload, instant AI diagnosis, and direct dentist sharing; cut average journey from referral to diagnosis from days to under 5 minutes.
• Implemented a human-in-the-loop retraining pipeline where approved annotations feed back into model training; each cycle improved mAP by ~3–5%.
• Built a Gemini LLM integration for real-time structured reports tailored for clinicians and patients, reducing post-diagnosis documentation to under 3 seconds.