Retrieval-augmented medical assistant that answers questions over user PDFs (textbooks, reports, guidelines) with concise, source-grounded responses.
ViLT-based VQA application that interprets an image and answers free-form questions, demonstrating end-to-end vision-language inference.
Ask natural-language questions over long articles and academic PDFs; returns focused answers and inline citations via RAG.
FLAN-T5 summarizer exploring zero/one/few-shot prompting to improve conversational compression quality and faithfulness.
Next-day close prediction with comparative baselines; includes EDA, feature engineering, model evaluation, and robustness checks.
Physics-informed neural network that solves the 1D viscous Burgers’ PDE by enforcing data and governing equations during training.
Minimal blog backend with JWT auth, CRUD for posts, password hashing, and SQLAlchemy models over PostgreSQL.
Demonstrates GitHub Actions + CML for automated training, evaluation, and artifact reporting in a TensorFlow workflow.