Research Engineer, Member Of Technical Staff
CurrentInformation Exploitation Lab.I evaluate, implement, and modify recent works in the field of artificial intelligence (deep neural networks are popular nowadays) to meet the varied needs of the defense community.A selection of unclassified work done/in progress:-Deep reinforcement learning in wargaming, cybersecurity, and AI safety (particularly distributional shift).-Person Reidentification (identifiying people across different cameras) using various modalities - appearance, gait, face - with deep neural networks (CNNs & RNNs) & large-scale people datasets. This was achieved by having neural networks learn representations of people that are invariant of pose, illumination, and background, but are affected by the targeted modality (e.g. appearance). Media coverage: https://www.straitstimes.com/singapore/room-to-work-on-robotics-and-ai-in-new-dso-complex-Adversarial attacks & defences on neural networks. Placed 25th out of 90 competitors in the 3rd development round of the Defence category in the NIPS Adversarial Attacks & Defences 2017.-Multi-band satellite image detection and segmentation with convolutional neural networks (e.g. U-net)-Geolocation for a classified project. Assessed the performance and fused geolocation algorithms originally meant for less extreme scenarios compared to the project tasked.Non-technical:-Encouraged students to take an interest in defense related research through outreach activities.-Shared lab's capabilities with external parties through talks held.-Guided students in short-term projects.
