Large-scale multimodal transformer architecture designed with custom attention mechanisms and optimized for efficient hardware use. Implements advanced cross-modal fusion, novel positional encoding, and memory-efficient inference patterns.
A unified transformer architecture for autoregressive text, image, and audio generation. Features breakthrough cross-modal attention mechanisms enabling seamless multimodal content creation with shared representations.
A transformer model that learns to show its reasoning steps before giving an answer. Implements chain-of-thought reasoning for transparent AI decision-making and improved interpretability.
Python framework for training and running transformer models using 2-bit (ternary) precision, achieving significant memory and compute savings while maintaining model performance.
A framework for experimenting with 4-bit floating point precision in deep learning models, exploring novel quantization techniques for efficient neural network inference.
Comprehensive comparison project analyzing 4-bit floating point vs 4-bit integer precision in deep learning, providing insights into performance trade-offs.
A deep learning model that classifies flower images into species using computer vision techniques and convolutional neural networks.
Web application for storing images and searching them by text or image using AI embeddings and vector similarity search. Features real-time search and semantic understanding.
Local application to upload documents and search their content by meaning, not just keywords, using semantic understanding and natural language processing.
Web tool that analyzes resumes and provides feedback using AI and NLP. Offers suggestions for improvement and keyword optimization for better job matching.
Tool for removing backgrounds from images using advanced AI models and computer vision techniques. Provides clean, professional results.
Research project exploring how large language models can manage and retrieve large amounts of context efficiently using novel attention mechanisms.
A modular AI system to automate all steps of CNC manufacturing, from design to real-time control. Integrates computer vision, path planning, and process optimization.
Comprehensive notes and code for designing large multimodal language models, covering architecture decisions, scaling strategies, and implementation details.
A simulator for estimating AI model inference speed on different NVIDIA GPUs, helping with hardware selection and performance optimization.
A tool for exploring GPU rack configurations and optimization, helping with datacenter planning and resource allocation for AI workloads.
An English-like programming language designed to make AI and ML workflows easier to write and understand, bridging the gap between natural language and code.
Small projects testing different AI models and agents for development tasks, exploring the future of AI-assisted programming and automation.
A comprehensive set of 60 Python programs covering various topics and use cases, from basic algorithms to advanced data structures and AI implementations.
Web project for visualizing AI concepts and algorithms, making complex machine learning ideas more accessible through interactive demonstrations.
Comprehensive web suite for visualizing different algorithms, helping students and developers understand algorithmic concepts through interactive animations.
Collection of modern web development design projects showcasing various UI/UX concepts and responsive design techniques.
Python project for generating and solving complex mazes using procedural generation algorithms and pathfinding techniques.
Collection of comprehensive reports analyzing various AI research topics, providing insights into cutting-edge developments and future trends.
Collection of synthetic datasets for training and testing AI models, covering various domains and use cases for machine learning research.
Document management tool with AI-powered search and organization features, enabling intelligent document retrieval and categorization.