Aasmeet Kaur

Architected a Retrieval-Augmented Generation pipeline to extract and answer questions from PDFs using FAISS vector search and transformer embeddings; engineered semantic chunking and similarity retrieval; delivered an open-source, API-free solution using local models for offline deployment; deployed an interactive Streamlit dashboard with source chunk highlighting.

Ongoing research project using the LC25000 histopathology image dataset. Engineered a multi-class tissue classification pipeline, implemented CNN and ResNet-based transfer learning, designed normalization and augmentation preprocessing, established evaluation using accuracy, precision, and recall, and applied feature visualization for explainable AI research.

Built an end-to-end machine learning pipeline on a 7,000+ sample telecom dataset; performed feature engineering and exploratory data analysis; trained and compared Logistic Regression and Random Forest models; improved recall using an ensemble approach; deployed a real-time Streamlit prediction app.

Developed a text classification system on 10,000+ news articles using TF-IDF vectorization and NLP preprocessing; benchmarked multiple classifiers and analyzed precision-recall trade-offs; delivered a reproducible pipeline from raw text ingestion to labeled classification output.