At GNDU Amritsar × India Meteorological Department, I designed and implemented a deep learning pipeline for meteorological prediction using BiLSTM and Bahdanau Attention. I worked across data preprocessing, feature engineering, model training, evaluation, and inference, and integrated LLaMA 3 to generate advisory bulletins.
I co-authored an accepted Indian National Science Academy publication on a multi-task cyclone prediction framework. The work included a literature review, benchmarking, and experimental validation.
I deployed MediAudit, a medical AI pipeline that uses computer vision for document OCR and NLP for safety auditing, drug-interaction checks, and structured lab-value analysis. Its FastAPI layer supports multi-model routing, validated output, bilingual responses, and audit history.
My projects also include a multi-agent AI news aggregator with RAG retrieval and a domain-specific fine-tuning workflow for Mistral and Llama. As a freelance AI Model Evaluator, I assessed coding agents on engineering tasks and reported on tool-calling and multi-step reasoning failures.
