Jetpack Joyride

This is a pure C++ game application that uses SFML for the graphics and user interaction. It uses Quad Tree for efficient collision detection by spatially partitioning game objects, a queue to manage power-ups and their activation sequence, and a BST to maintain the leaderboard efficiently. It also employs OOP principles for organized design, handling enemy and player logic through separate classes. It uses inheritance in the power-up class hierarchy to manage various power-ups efficiently.

Civic IQ Verifier

Developed an automated identity verification system for Aadhaar and PAN cards using YOLOv8 for object detection, Tesseract OCR, Real-ESRGAN, and face recognition. The system extracts structured fields from documents, decodes secure QR codes, and performs face matching with live selfies. Enhanced QR code quality using Real-ESRGAN and custom deblurring techniques. Integrated DigiLocker mock responses for cross-verification of OCR data. Designed as an easily integrable Python package for automated document verification workflows. Future plans include expanding document support and improving security features.

Instagram ChatBot

Developed an automated Instagram chat bot using Puppeteer and Node.js to detect and respond to messages. The bot identifies unread messages, filters abusive language, and replies with predefined responses. Incorporated a Hugging Face model (Mistral-7B-Instruct) fine-tuned using PEFT on personal WhatsApp chat data to generate Hinglish-style responses that reflect my conversational tone.

CodeAct-Style SQL Agent

reasoning agent built in Python that uses Mistral-7B-Instruct via the Hugging Face API to convert natural language questions into executable SQL queries. Follows the CodeAct and ReAct approach to reason step-by-step, generate Python code dynamically, execute it on a SQLite database, and refine queries through iterative self-correction.