Machine Learning Engineer, Amazon Web Services

Created an Agentic AI system with anti-hallucination guardrails, intent classification, task decomposition, orchestration caching, and a novel dynamic tool generation capability with caching that fetches from a curated vectorized knowledge base of AWS datacenter best practices and live AWS datacenter build data. This allows datacenter professionals to interact with an intelligent assistant for their daily tasks, reducing an estimated 40% of the time they spend on manual analysis. Built with Amazon Bedrock, Open Search Serverless, Strands, and MCP, with support for multimodal models like Claude, Nova Premier, and Command R+.
Developed secure, serverless, ML-based predictive algorithms for project management schedule evaluation, risk identification, and delay prevention, utilizing Amazon Textract OCR, Lambda functions, DynamoDB, S3 bucket encryption, Simple Queue Service, Docker, IAM auth, Amazon Redshift, and API Gateway. Saved $75M+ in mitigation costs, per Amazon management estimates.
Integrated with 10+ internal tools, creating a seamless user experience within the larger AWS ecosystem, by collaborating with global teams in APEC, AMER, and EMEA regions. Contributed to an internal tool with a user base of 18,000 users.