Guru Varshini B

I’m an Agentic AI Engineer with two years building production multi-agent and RAG systems, plus hands-on frontier model evaluation. I design HITL-enforced agentic pipelines and implement AWS Bedrock Guardrails for prompt injection blocking and PII protection, then evaluate LLM output quality using structured rubrics and LLM-as-judge frameworks.

I’ve shipped multi-tenant platforms and agent workflows that measurably improve retrieval and reliability—like a self-correcting Agentic RAG system with labeled evaluation harnesses and strong gains (recall@k, MRR) plus out-of-domain abstention. At Turing (contracted to Anthropic, Microsoft, and ServiceNow), I lead quality calibration, build failure taxonomies for agentic tool-use errors, and support RLHF and multi-tool agentic workflow training across enterprise domains.