Project Blinkem - Optical channel communication for Mixed Reality Systems

Project Blinkem is a novel approach to visible light communication, which aims to create a low-cost, scalable, and secure framework using existing technologies. The project proposes encoding data into Infrared LEDs, which can visually communicate with high-resolution image sensors on HMDs as well as a high-speed optical module. By leveraging the security benefits of visible light communication, Project Blinkem offers an inherently more secure method of communication, which can be valuable for a range of applications, including virtual reality environments and the internet of things.

Machine Learning for Speech Enhancement

The core value of Nureva is to provide reliable and easy-to-use audio-conferencing products that offer a good user experience and maximize productivity. The unique Microphone Mist™ technology unlocks new possibilities to pick up audio from anywhere in the room. One subtopic of this research is to investigate ML as it relates to sound event localization and detection so we can track when and where a person is talking in the room. As a result, this can allow virtual microphones to be activated near the talker while attenuating noises in the room.

Model compression and inference optimization for UAV’s companion computer

With the evolution of computer vision methods in recent years, more and more scene-understanding methods have been proposed. Meanwhile, SOTI also designed and developed several scene-understanding algorithms. However, due to the limited computing capabilities of UAVs, all these algorithms need to be compressed to use fewer resources. This project will deliver a model compression and deployment pipeline.

Software System Error Detection and Resolution

The software development process is a lengthy process and an area where most companies spend a great amount of capital. Approximately half of this time developers are spending on fixing bugs in their code. Faulty software is difficult to identify both in location and reason. After finding a bug, it takes even more time to identify the correct solution to the problem. In this internship, we propose to create a novel approach that identifies topics and relationships between bugs in code.

AI-driven Predictive Models and Consumer Insight for Trade Optimization Improvement

The proposed project is to develop AI strategies to provide precision marketing through consumer segmentation and recommender systems, as well as to promote events that shall meet various business goals for retailers and Unilever. Successful outcomes will feed into an On-Demand AI Engine aimed at improving consumer engagement and pricing strategy in the consumer packaged goods sector.

AI Based Script Builder for Web Payments

To conduct research and provide a feasible solution to create an AI-based script builder to automate the company’s pay-by-web transaction process. The pay-by-web transaction process includes many steps. These are steps such as extracting data from different sources and monitoring email inboxes as well as verifying payment information. The process also involves identifying the correct supplier websites and submission of the complete payment transaction. The project aims to provide an AI script that automates this process to make it more streamlined and efficient.

Using multi-modal data and self-supervised approaches for machine learning in healthcare

This research project aims to address the growing interest in predicting clinical outcomes using machine learning
(ML) approaches applied to Electronic Medical Record (EMR) data. The primary objective of this study is to
develop representations of both EMR and text data found in medical notes using current state-of-the-art ML
techniques. In particular, this research proposes to leverage self-supervised learning techniques to learn dynamic
representations. By doing so, the research aims to improve the prediction of clinical outcomes.

Driver Behaviour Analysis Using Accelerometers on Android Devices

The objective of this project is to develop a software solution that can analyze the accelerometer data on android devices and report certain metrics of a moving vehicle related to road safety and driving conditions. The reports can then be used for immediate call for action in case of detecting an emergency or collecting the statistical data over a longer time period for assessing the general driving behavior.

Data Communication Optimization between Mobile Devices and Servers

The proposed research project aims to improve the way data is transmitted between mobile devices and company servers. This will improve the speed and security of file transfers, data synchronization, application deployment, and distant management of mobile devices. The intern will collaborate with experts from the partner organization who will offer guidance and support in this research to identify new techniques that can be used to reduce the amount of data being transmitted, while also ensuring that the data is secure.

Development of a distributed framework for deep learning models

Layer 6 powers the AI use cases for a variety of banking and financial applications at TD Bank. The goal of the research project is to improve the AI engine by having the training more efficient and distributed among a variety of clusters. The AI engine will allow models to be trained faster and with more optimal performance of the models. An improved AI engine can help deliver better machine learning models to over 25 million customers that rely on TD Bank for their financial decisions.