Open to opportunities

Nancy Sakhiya

@nancysakhiya

Message

Entry-level AI and ML researcher building explainable and production RAG systems.

India

Message

What I'm looking for

I want to build explainable ML and production RAG systems—benchmarking SHAP/LIME, shipping fast APIs and reliable data pipelines, and deploying user-focused AI. I enjoy reproducible work and clear collaboration.

I’m an entry-level AI/ML researcher focused on explainability and practical delivery. In my current Research Intern role in Explainable AI (XAI), I benchmark SHAP and LIME on tabular ML models by evaluating explanation fidelity, sensitivity, and feature attribution consistency under input perturbations.

I also build production-grade multimodal intelligence systems and applied ML apps. My DoCQA multimodal RAG project uses FastAPI and Next.js with OCR (Tesseract/PyMuPDF), Docker Compose deployment, and an embedding workflow with nomic-embed-text stored in pgvector on Supabase, grounded with Llama 3 and real-time SSE streaming; I’ve built end-to-end forecasting and sentiment dashboards with LSTM + FinBERT deployed on Hugging Face Spaces, and I co-authored a published study introducing a hexagonal pixel representation that improves segmentation performance.

Experience

Work history, roles, and key accomplishments

Current

Research Intern — Explainable AI

Current

IIIT Vadodara

May 2025 - Present (1 year 2 months)

Conducts a comparative study of SHAP and LIME for post-hoc explainability on tabular ML models, evaluating explanation fidelity, sensitivity, and feature attribution consistency under input perturbation. Performs hyperparameter tuning and model evaluation (Random Forest, XGBoost, baselines) on the UCI Abalone dataset and visualises global/local feature importance to document benchmark findings.