A RAG web application that enables context-aware, multi-turn conversations over uploaded PDF documents. It uses Streamlit, LangChain, and Gemini, with a focus on secure API key management and modular execution state.
A RAG system that answers questions by analyzing entire YouTube playlists and citing precise video timestamps. It handles Hinglish queries using BAAI/bge-m3 embeddings, stores them in Qdrant, and runs asynchronous video transcription via Faster-Whisper.
An AI physiotherapy app that gives real-time feedback on exercise form. It uses MediaPipe to track body landmarks, calculates joint angles to count repetitions, and corrects user posture on the fly.
A production-grade web application that predicts home prices using a Scikit-learn Linear Regression model. It features a Flask REST API, an Nginx reverse proxy, and a responsive frontend for users to query property values.
An offline-first payment backend that settles UPI-style transactions without internet connectivity. It relies on a simulated Bluetooth mesh network and hybrid RSA/AES cryptography, ensuring idempotency and secure offline ledgers.
A C++17 deep packet inspection engine that analyzes PCAP captures and reconstructs network flows. It classifies application-layer traffic using TLS SNI and HTTP Host inspection, supporting both single-threaded and multi-threaded architectures.
A custom C++ vector database implementing HNSW, KD-Tree, and Brute-Force search. It serves as the backend for a fully local RAG pipeline using Ollama, complete with a REST API and a frontend for 2D PCA visualization of the vectors.
An interactive AI chatbot that represents a job candidate, grounded entirely in their parsed PDF resume to prevent hallucinations. It turns a static resume into a conversational agent using a FastAPI backend, Pydantic schemas, and Gemini.
A machine learning pipeline for predicting property prices in Mumbai. It includes a Random Forest model, a Streamlit web app, a Tkinter desktop GUI, and an SQLite database for persistent records.
An automated resume scoring tool that surfaces candidate skill alignments and gaps. It processes resumes in batches with caching and resilient API calls to Gemini, presenting the results through a Streamlit interface.
A financial fraud detection system powered by an XGBoost model. It handles severe class imbalance using SMOTE and temporal feature engineering, and provides clear, interpretable reasons for its predictions using SHAP values.
A time-series forecasting tool that predicts renewable energy generation from solar and wind sources. It uses rolling window features and lag variables to model temporal trends, surfaced through an interactive Streamlit dashboard.
A computer vision classifier that identifies sports celebrities. It uses OpenCV Haar Cascades and Wavelet Transforms to extract robust facial features, classifying them with models like SVM and Random Forest behind a Flask API.
A stateful AI restaurant ordering system powered by LangGraph and Gemini. It handles multi-turn conversations using a FastAPI backend with MemorySaver, calls tools to query the menu, and serves users through a Streamlit frontend.
An automated job-search pipeline that pulls roles from public ATS APIs, filters out noise, and scores them against a resume using a two-stage LLM screener. It then automatically drafts tailored application kits and cover letters for the best matches.
An intelligent learning analytics platform that generates dynamic assessments and tracks behavioral telemetry. It uses RAG LLMs, a Supabase PostgreSQL backend, and a Streamlit interface.
To associate your repository with the deep-analyzed topic, visit your repo's landing page and select "manage topics."