Work Experience

• Architected a large-scale medical document analysis platform for Gilead Sciences using AWS Bedrock and Databricks, processing 3k+ clinical documents monthly and reducing manual research analysis effort by 70%.

• Created a RAG-based Knowledge Assistant indexing 30k+ medical documents and 2M+ vector chunks, enabling low-latency semantic retrieval across enterprise research datasets.

• Processed 100GB+ structured and unstructured medical datasets weekly through Databricks ingestion, embedding, and vector indexing pipelines.

• Orchestrated multi-agent reasoning pipelines with AgentCore and Bedrock to automate anomaly detection across 1k+ clinical reports weekly, shortening report triage turnaround time by 60%.

• Integrated OpenTelemetry tracing and LLMOps evaluation pipelines across 20+ distributed AI services, reducing debugging and root-cause analysis time by 70%.

• Browser Automation: Architected a high-throughput automation infrastructure using Playwright and AI-driven heuristics to execute 30,000+ daily workflows, successfully replacing high-volume, repetitive manual operations with intelligent agents.

• Investment Assistant: Spearheaded the development of a stateful AI Investment Assistant serving 100,000+ users, designed to augment Relationship Managers (RMs) by synthesizing client knowledge and regulatory guidelines in real-time.

• Smart AI Cursor: Designed the "Interaction Brain" for the flagship Smart AI Cursor, building a proprietary intent-recognition engine that translates raw, unstructured user inputs into precise, executable workflows.

• Content Creation: Built an autonomous end-to-end content pipeline for assessment generation, orchestrating the entire lifecycle from creation and analysis to validation and seamless ingestion into client ecosystems.

• Engineered a centralized orchestration console to trigger and monitor 30+ existing Azure workflows, creating a "single pane of glass" that reduced manual tracking effort by 50%.

• Architected a fault-tolerant ingestion pipeline processing 10k+ daily records by implementing Message Queues for load buffering and Write-Ahead Logging (WAL) to guarantee data durability.

• Developed high-performance backend services in FastAPI and Go with real-time WebSocket interfaces, maintaining 99.9% uptime while providing instant visibility into asynchronous tasks.