VS CODE EXTENSION, Developed a standalone PyQt based GUI extension for VS Code integrating Competitive Programming Helper (CPH) with LeetCode, using a QScintilla editor and real-time output for Python and C++, eliminating the need for browser/terminal., Automated problem and test case fetching via the LeetCode API with filters for difficulty, tags, and company; designed a clean, intuitive interface for an efficient coding workflow., Tinkering Lab, IIT Roorkee Using Physics Informed Neural Networks (PINNs) to solve Differential Equations, Built a tsunami prediction and early warning system using Physics-Informed Neural Networks (PINNs) trained on Shallow Water Equations and NLSE (Nonlinear Schrdinger Equation) with custom residual loss for accurate coastal wave simulations., Modeled wave height and velocity over space-time using PyTorch-based deep neural networks, simulating real and imaginary components to capture nonlinear wave behavior., IIT Roorkee Song recommender for Instagram story, Built a Flask-based AI app that recommends songs by analyzing the mood and aesthetics of uploaded images using BLIP for captioning and Gemini VLM for refined visual understanding., Matched SBERT embeddings of captions with precomputed song vectors via cosine similarity, with filters for artist and language preferences; integrated into a responsive React and Tailwind CSS UI., Quantum Computing Group, IIT Roorkee Generative Text for Customer Support Automation, Developed an interactive Streamlit chatbot combining retrieval-augmented generation and rule-based logic, with modular services like FAQ automation, mood-based music recommendation, and contextual engagement., Used a CSV-based FAISS VectorDB with LangChains RetrievalQA for semantic responses, and integrated Google PaLM API with ConversationBufferWindowMemory for context-aware, multi-turn dialogue; deployed via a custom CSS and sidebar-driven UI., Consulting & Analytics Club, IIT Guwahati Stock Sentiment Analysis using Machine Learning, Built a sentiment-driven trading signal system for Ethereum by combining historical price data with real-world news sentiment., Implemented web scraping and preprocessing pipelines (Selenium, BeautifulSoup, webdriver-manager) to collect and clean news and market data; extracted sentiment features using TextBlob and VADER., Merged sentiment metrics with financial indicators (returns, price change, P&L) to engineer a labeled dataset and trained ML models (Logistic Regression, SVM, Random Forest) for actionable market predictions., Finance Club, IIT Roorkee