Open Projects for Master Students

Send me your CV to chang.gao at if you are interested in an open project or if you want me to create a project for you.

This project focuses on creating a chatbot using the innovative RWKV RNN-LM (Recurrent Weighted Key-Value Memory Network Language Model) architecture and PyTorch framework. The goal is to develop a chatbot similar to ChatGPT with human-like text generation capabilities. Additionally, the project explores model compression techniques, such as quantization and pruning, to optimize the language model for deployment on an NVIDIA Jetson Orin Nano, a compact AI computer designed for edge computing applications. Key tasks include researching the RWKV RNN-LM, setting up the Jetson Nano, preprocessing training data, implementing model compression techniques, and creating a user-friendly interface. The ideal candidate has experience with Python, PyTorch, deep learning, and embedded systems.

Audio denoising is a critical component in modern voice-controlled devices, digital assistants, and other audio-centric applications. The increasing demand for high-quality audio and low-power consumption necessitates the development of efficient and effective denoising techniques. This master's project aims to design, implement, and evaluate a neuromorphic hardware accelerator optimized for audio-denoising tasks, taking advantage of the parallelism, low power consumption, and real-time processing capabilities of neuromorphic computing. The project will involve a thorough investigation of existing audio denoising techniques, neural networks training and optimization, and neuromorphic hardware architectures. The primary goal is to develop a hardware-software co-design approach that seamlessly integrates the neuromorphic accelerator with existing audio processing systems, ensuring compatibility and ease of deployment.

Check this paper out: 2303.09503.pdf (

The rapid growth of wireless communication and the increasing demand for high-speed data transfer has driven the development of 5G networks. A key challenge in 5G systems is improving the energy efficiency of wireless transmitters, especially in the context of mobile devices and base stations. High Peak-to-Average Power Ratio (PAPR) in Orthogonal Frequency Division Multiplexing (OFDM) signals can lead to decreased efficiency in power amplifiers, causing increased power consumption and reduced battery life in mobile devices. This master's project aims to apply deep learning techniques to reduce the PAPR in 5G OFDM signals. By doing so, the project seeks to enhance the energy efficiency of wireless transmitters in mobile phones and base stations, thus improving overall system performance and user experience.

Always-on keyword spotting on ultra-low-power integrated circuits is a hot topic in chip design for artificial intelligence systems. Previous work used spiking neural networks (SNNs), Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), but the state-of-the-art accuracy was recently achieved by transformers. However, transformers have a large memory footprint and many arithmetic operations; thus, it is difficult to be employed on tiny embedded systems with scarce resources. In this project, you will work on developing algorithmic methods to compress the model of transformer neural networks to reduce its memory and arithmetic cost. 

Cochlear implants (CI) [1, 2] are miniaturized biomedical devices that can help deaf people perceive sound or help hearing loss patients understand speech better. The CI has an in-vitro module attached behind the ear and an in-vivo implant surgically placed under the skin. The quality of CI output signals degrades in noisy environments and relies on Speech enhancement (SE) systems to enhance its performance. Neural network-based SEs [3] achieve state-of-the-art performance but are expensive to deploy on CI with a limited power budget. In this work, you will build a small-footprint NN-based SE system that can run on embedded devices [4, 5, 6] in real-time.