The fellowship mainly investigates an analysis of the state-of-the-art approaches, design and implementation of cutting-edge deep neural network models to be used on a mobile platform. It explored ways to optimize the deployment of these machine-learning models for prediction tasks on the mobile devices which requires energy efficiency and accuracy.
Currently, some public and private organizations have implemented various identification verification solutions to manage identity authentication. The idea of using a third-party identity provider (IdP) to access a relying party (RP) is not new, and both RP and IdP have their benefits as they can only be connected once in a federated identity ecosystem. While the deployed identity brokerage system has provided participants with great utility, it was pointed out that the principles they designed had several security and privacy gaps.
Global service providers in highly regulated financial sectors must accommodate an ever-changing, sometimes competing, landscape of regulatory and business concerns. This project seeks to define a technology infrastructure design that supports current and anticipated data privacy and data residency concerns, making it possible to keep data within borders while still facilitating collaboration across those borders. Consumers are increasingly aware of the collection of their private data, but are often unaware of cross-border movement of their data.
Global service providers in highly regulated financial sectors must accommodate an ever-changing, sometimes competing, landscape of regulatory concerns. This project seeks to determine a reasonable path forward in technology design and adoption to accommodate current and anticipated data privacy and data residency concerns, making it possible to keep data within borders while still facilitating collaboration across those borders. Consumers are increasingly aware of the collection of their private data, but are often unaware of cross-border movement of their data.
Ecobee is a home automation company that makes thermostats for residential and commercial use. Smart lights are the second-most desired home automation devices after thermostats, but being able to remotely control the lights is only a small part of the vision. ecobee light switches have embedded microphones, a speaker, and far-field voice technology that allows the user to control the lighting, thermostat, and other smart home products. One major goal of this project is to create an artificial agent, which monitors and controls the light switches.
The Pungle payments-as-a-service platform delivers low cost, real-time, friction free business disbursements, peer-to-peer (P2P) transfers, and B2B supplier payments. Pungles mission is to enable businesses with a digital payments platform that provides real-time disbursements and transfers. The problem that arises with digitization of business payments is higher risk for fraud due to its electronic nature. Therefore, theres a need to be absolutely certain that both the sender and recipient of payments are the intended parties and that there are no anomalies in payment volume and frequency.
Nowadays a corporations public image plays a major role in that companys decisions and financials. This project involves predicting fraud and errors within the financial statements of publicly traded companies. The goal is to incorporate information such as press releases and industry media coverages to provide an insight to these companies under audit and their industries.
Serving as the most widely-used body part for communication, hand is a very important tool for human to interact with the world. Especially with the continuing development of virtual reality and augmented reality, hand pose information has gradually become an indispensable component for improving users experience in interacting with computing devices. Therefore, this project aims at achieving hand pose reconstruction based on capacitive sensing technology using machine learning algorithm.
Addictive Tech Corp is a fast-growing ad-tech company. They use real-time advertisement bidding software which is massive and sophisticated. The actual dynamics involved in any given bid are complex and hard to predict. This makes writing test logic for such a system cumbersome and catching all corner cases next to impossible. Because of the scale of operations, understanding the environment in which the bidding software operates is difficult. This is problematic as such software needs to be highly optimized to be competitive.
Artificial intelligence, especially Machine learning algorithms, plays important roles in building recommendation systems and promotional forecasting systems for retailers. However, training a machine learning model requires the choice of a number of significant features and requires tuning a large set of configurations. Therefore, it takes a long time for humans to find the optimal configuration for one or more predictors. However, the predictive performance of existing automated tuning models is not as good as manually tuning. Besides, the approach cannot be applied to more than one model.