Kunal Aggarwal

Built GraphIQ, a production-grade conversational AI platform leveraging LLMs, MCP, and agentic framework, reducing manual meeting coordination time by ~80% via Google Calendar integration. Designed an agentic workflow engine with LangGraph, Celery, and FastAPI supporting asynchronous multi-agent operations and improving task parallelism by ~35%, reducing response latency across workflows. Developed a speech-to-speech system integrating ASR, TTS, and RAG components, achieving ~95% speech recognition accuracy and enabling real-time, low-latency voice-based interactions with sub-2s response time. Fine-tuned Mistral-7B and OpenAI 4o-mini models on a specialized medical dataset, achieving 85% accuracy for extracting disease specific ICD and CPT codes. Served as the sole AI/ML Engineer in the Quick Response Team (QRT), delivering 10+ rapid AI PoCs for client demos and internal showcases, reducing concept-to-demo turnaround time by 40%.

Worked on ML-driven data analysis and healthcare model development; improved accuracy by 30% through feature engineering and deployed models via Flask API.