uTensor is an open-source framework that pioneered enabling efficient machine learning on resource-constrained devices. It is designed to be portable, lightweight, and easy to use.

A project to bring machine learning to the BBC Micro:bit, a popular educational hardware coding platform. The project aims to make machine learning more accessible to students and educators.

Pre-trained a 1.58-bit Llama-inspired LLM that can fit all its parameters in 32 MB of memory. It is supervised finetuned (SFT) and RLHF aligned for specific use-cases. The project pushes the boundaries of generative models on resource-constrained devices.

This exoskeleton enhances the quality of life for elderly individuals by providing support for walking and standing. Using sensor fusion and heuristic control algorithms, it anticipates and assists with movement. The prototype was developed in just three weeks as part of an award-winning hackathon project.

This project uses an U-Net based Generative Adversarial Network (GAN) to generate new Gundam designs. The GAN is trained on a dataset of existing Gundam designs and can create new designs that are both novel and realistic.