Ms Sanjay

AI Engineer with 2 years of experience shipping production Agentic AI, Retrieval-Augmented Generation (RAG), and

LLM-integrated systems on AWS Bedrock and Claude. Built multi-agent workflows with tool/function calling, retries, and

Human-in-the-Loop approval gates across enterprise data and healthcare platforms, including a customer-data normalization system that cut manual cleanup effort by 50–70%. Focused on grounded, guardrailed LLM outputs, agent evaluation, and end-to-end delivery from architecture through production.

• Led end-to-end technical design, development, integration, and delivery of AI-powered backend solutions across two production projects, owning architecture, implementation, troubleshooting, performance tuning, production support,

and LLM observability (Datadog) for a team of 5–8 engineers and stakeholders.

• Architected multi-agent Generative AI workflows on AWS Bedrock and Claude with tool calling, memory, and retry/fallback logic, enabling agents to make context-aware decisions rather than execute static prompt chains.
• Built Human-in-the-Loop approval gates into agent decision paths for enterprise-scale customer data workflows, ensuring zero unreviewed automated pipeline changes reached production.
• Developed intelligent automation covering document processing, data extraction, target-schema mapping, customer-data normalization, agent routing, and API-driven pipeline updates via FastAPI, processing hundreds of records per day.
• Implemented agent evaluation checkpoints and guardrails — including PII detection and irrelevant-query filtering — to validate LLM-generated outputs for accuracy, consistency, and safety, reducing invalid/unsafe outputs reaching production by 50%+.