Abhijay Gupta

At Netision Technology, I built GreenMind BI, a local multimodal enterprise BI prototype that combines PDFs and spreadsheets with validated analytics and source-grounded answers. I engineered a pipeline that turns complex questions into typed evidence requirements, executes preservation-validated read-only SQL, and retrieves document evidence for a transient analytical workspace.

I promoted the verified workflow behind a fail-safe local feature gate, achieving 20/20 controlled answer-quality and integrity cases, 60/60 grounding cases, and zero incorrect acceptances across 38 generated-schema tests.

Alongside my B.Tech studies in Computer Science and Engineering (AI), I’m developing a pip-installable RAG reliability SDK for claim-level evidence verification, intervention-based audits, and reproducible held-out evaluation. I also reproduced a leakage-controlled murmur detection baseline on the CirCor DigiScope dataset, reaching approximately 80.1% weighted accuracy under patient-level evaluation.