Projects
Mini GPT: QA Model from PDF
Architected a QA pipeline leveraging RAG and vector search to extract insights from PDF documents. Boosted document comprehension accuracy to 90% by embedding Qdrant vector stores with HuggingFace models.
Read MoreCredit Card Risk Modeling
Developed a binary classification model to identify high-risk credit card applicants using demographic and transaction data, improving default prediction accuracy to 93%. Applied SHAP analysis to enhance interpretability and transparency in financial decision-making models.
Read MoreMilk Quality Prediction
Engineered supervised learning models to classify milk quality, automating testing for 1,000+ samples. Reduced manual testing time by 70%, streamlining quality assurance processes for potential use in dairy production.
Read MoreReal-time Air Quality Prediction
Trained time-series forecasting models to monitor air pollution trends using LSTM and SARIMA. Delivered 85% prediction accuracy on unstructured data, enabling proactive planning in high-density urban regions.
Read More