Projects

F1GPT is a domain-focused AI assistant built with Next.js and powered by GPT-5 through the AI/ML API. It uses a Retrieval-Augmented Generation (RAG) pipeline that scrapes and indexes authoritative Formula One sources — such as Wikipedia and the official F1 site — into a vector database (DataStax Astra DB). When a user asks a question, the system embeds the query, retrieves the most relevant chunks, and streams a GPT-5 answer enriched with this context. The chatbot is deployed with a sleek Formula One-themed UI, enabling real-time streaming responses. Its technical stack includes Next.js 15, Vercel AI SDK, LangChain text splitters, Puppeteer for scraping, and Astra DB for high-performance similarity search. Value & Use Cases: - Fans can quickly look up race results, driver stats, or rule explanations. - Journalists can explore historical data and technical regulations faster. - Teams & Analysts can leverage a conversational layer over structured F1 data. By combining GPT-5’s reasoning power with RAG, F1GPT reduces hallucinations and delivers trusted, up-to-date answers. This positions it as a unique sports-focused AI assistant, showing how AI can enhance fan engagement and knowledge accessibility in high-data domains like Formula One.

The Final Year Design Project aims to improve object grasping and manipulation in warehouses. It features an autonomous robotic arm that efficiently retrieves items from any defined place using dual monocular cameras for 6D object pose estimation and techniques involving robotics and machine learning for grasping strategies that includes kinematic modeling and inverse kinematic modeling of the robotic arm.

The SmartScan Billing Assistant is a computer vision-based project designed to automate the process of calculating the total bill by detecting products and adding their prices using a camera, similar to a cashier. This project uses the YOLOv8 model for object detection and tracking. This is the MVP of the main project I did for a supermarket where I implemented the SmartScan billing system on more than 20 products.

Facial Emotion Detection using YOLOv8 is a Flask web application designed to detect facial emotions in real time. Leveraging the power of the YOLOv8 model, this project offers a robust solution for identifying a range of emotions directly from images or webcam feeds.

Human Sense Ads is a project designed to detect humans, extract their features, and generate personalized messages using a language model (LLM) to grab their attention towards advertisements displayed on a screen.