Looking for a job

Vaidehi User

@vaidehi_amrutkar

Message

I build and evaluate AI systems that expose unsafe agent behavior and improve reliable model performance.

India

Message

What I'm looking for

I'm looking to build and evaluate reliable AI systems, especially work involving model safety, agent evaluation, fault injection, RAG, and measurable system performance.

I'm building AI evaluation systems that surface silent failures in autonomous agents. In Crucible, I created a fault-injection harness with guarded tool calls and LLM-as-judge classification; guarded agents handled 40% of injected faults safely versus 0% for unguarded agents.

I've also run controlled ML comparisons for coronary artery calcium segmentation across six architectures, achieving a 0.8820 Dice score and 3.8× faster inference than baseline. For NeuralScout, I designed an offline candidate-ranking pipeline that processed 100K candidates in 2.5 minutes and was validated by 161 automated tests.

At ARG Supply Tech, I built a Hybrid-RAG system over supply-chain datasets, reduced manual analysis effort by about 40%, and improved query efficiency by 30%. I also contribute to nnU-Net and CNCF Jaeger, improving medical-imaging preprocessing and distributed tracing telemetry.

Experience

Work history, roles, and key accomplishments

Generative AI Intern

ARG Supply Tech Pvt. Ltd.

Jan 2026 - Mar 2026 (2 months)

Built a Hybrid-RAG indexing and query system using Python, LangChain, and ChromaDB over large supply-chain datasets. Developed backend API using FastAPI and React, tuning embedding-based semantic retrieval to 2-4 second query latency. Automated complex data mining workflows, cutting manual data analysis effort by ~40% and improving query efficiency by 30%.

Python Lang Chain ChromaDB fastAPI React RAG Data Mining