Jagpreet Singh

Generative Pretrained Transformers

The project focus on Training a Generative Pretrained Transformers(GPT) aka foundation model from scratch using plain text. The foundation model than can be finetuned on a specific task like text generation, summarization, translation, and more. The project leverages the power of GPT-2 architecture, a state-of-the-art language model, to generate human-like text based on a given prompt.

Disclaimer This project is intended for learning purposes. Training a production-level LLM requires extensive computational power and data. Please conduct thorough research before implementing this in any production environment.

Text to SQL

This project involves finetuning Gemma 2 model, a Pretrained Transformer to generate SQL queries from natural language text. Integrated the model with SQLite to utilize schema information, providing crucial context for enhanced query prediction. This tool aims to simplify database interactions for non-technical users, enabling them to retrieve and manipulate data using natural language commands.

Document Reader

Introducing the Document Reader Chatbot App—powered by OpenAi GPT3.5-Turbo a Large Language Model, with a Chainlit GUI. Users can upload text/PDF files, saving time by interacting with AI for content comprehension. The chat interface enables seamless Q&A, extracting relevant info. Prioritizes security, scalability, and user feedback. Redefining document interaction for efficiency and personalized convenience.

American Sign Language

The "Hand Sign Detection with TensorFlow Lite and Flask" project combines computer vision and web technology to perform real-time hand sign detection. This application captures video frames, employs a pre-trained TensorFlow Lite model for sign recognition, adds unique frame identifiers, and showcases the results through a user-friendly Flask-based web interface. Leveraging powerful libraries like OpenCV and MediaPipe, this project exemplifies the synergy of machine learning and web development, enabling intuitive communication through hand signs.

Alzheimer's Diagnostic with Computer Vision

The aim of the project is to detect Alzheimer's disease at early stage using Open Access Series of Imaging Studies(OASIS) brain data set. This project involves using implementing machine learning techniques and exploring different algorithms and methods to accurately detect the disease through the given data. This model will detect change of structure is specific part of brain and abnormalities that leads to Alzheimer's disease.