Salary Prediction using Regression
Developed a salary prediction model using Python, Scikit-learn, Pandas, and GitHub. Cleaned and preprocessed data using label encoding and duplicate removal; compared Linear Regression, Random Forest, and Gradient Boosting; and implemented a Voting Regressor ensemble.
NyayaGPT – RAG-Based Legal AI Assistant
Developed a Retrieval-Augmented Generation system mapping natural-language crime descriptions to relevant BNS/BNSS Acts. Used embeddings and similarity search across 1,500+ indexed law sections, improving retrieval accuracy to approximately 85% and reducing response latency from approximately 3.8 seconds to 2.1 seconds.
Email Reply Suggester
Built an LLM-powered email reply system using few-shot prompting, Llama 3.3 70B, and Groq. Developed an evaluation pipeline measuring correctness, factuality, tone, and conciseness, with hold-out evaluation and human-versus-LLM scorer validation.