Agentic AI, RAG, AI Evaluations, and Data. Co-inventor on “Self Improving Assistant: A Dynamic Framework for Task Learning and Adaptation,” a patent-pending agentic AI framework that autonomously learns, adapts, and scales new functionalities.

Led AppleCare's Generative AI initiative, securing executive buy-in and delivering a production system that supports thousands advisors with millions of weekly responses. Architected a scalable generative AI solution using MoE models, fine-tuning, RAG/personalization, and real-time inference optimizations. Delivered an 8× latency reduction while achieving a 85% measured performance improvement. Shaped AI strategy at the executive level, modernized conversational platforms, and built a top-tier 24-person ML team.

Led cross-functional ML initiatives at Apple, leading a team of 9 and driving NLP advancements, geo expansion strategies, and data annotation tools. Reduced model expansion costs by 30%, effort by 80%, and improved data-driven accuracy by 12.5%. Spearheaded high-impact projects, mentored junior engineers, and elevated Apple’s internal ML community through conferences, tools, and events.

• Led effort to increase coverage of team’s core machine learning product by 9% and classification accuracy by 12.5% through various application level and model improvements.

• Cut training time of a production deep learning model by ~90% through rethinking/rewriting entire machine learning pipeline and model training process. Simultaneously increased accuracy of model by 10%.