Leading the platformization of GenAI at Coinbase, architecting a scalable, multi-agentic AI platform that enables developers to build advanced applications with grounded, cost-optimized, and latency-efficient responses. Spearheading out-of-the-box solutions that integrate multi-cloud LLMs, heterogeneous knowledge sources (enterprise data, web, semantic stores), and open-source agentic frameworks (LangGraph, CrewAI) to accelerate AI innovation for both internal and high-traffic, end-user-facing applications.

Architected and led the development of Coinbase’s GenAI platform, deploying scalable, multi-cloud AI solutions (AWS Bedrock, GCP Vertex, Azure OpenAI) that power 30+ mission-critical applications across the firm. Engineered seamless integration of multi-cloud LLMs and RAG systems while optimizing vector database scaling, prompt caching, fine tuning, guardrails, evaluation frameworks, multi-modality, and multi-agentic systems - delivering $40M+ in estimated cost savings.

Optimized real-time aggregation of frontend and backend events across all Coinbase services, feeding a time-series model to generate scaling signals for AWS Auto Scaling Groups, ensuring efficient CPU utilization and resource optimization.

Enhanced risk scoring for millions of BTC/ETH transactions per hour, leveraging Graph Neural Networks (GNNs) to detect fraudulent addresses and trigger real-time alerts for Coinbase customers.

Spearhead the design and development of a knowledge graph driven platform that extracts, inter-connects and stores structured knowledge from heterogeneous sources ranging from textual regulatory filings to structured market data. Contributed to designing and scaling a firm wide ML and document extraction platform that allows developers to create custom content extraction pipelines that embed complex ML models and can process a diverse format of financial documents

Designed and delivered production platform with search and extraction functionalities on regulatory documents.