David Wang Email & Phone Number

Singapore

Singapore

• Developing mutli-modality machine learning models combining video and text for enhancing the performance of TikTok video deduplication systems.

Singapore

• Adapted state-of-the-art generative diffusion pipelines for iOS using SwiftUI and Bazel, leveraging enhancements with CoreML, MPS, and Metal to achieve native on-device functionality with platform-specific optimization. • Spearheaded the project development, integrating insights from weekly analyses of cutting-edge research papers to enhance team expertise and guide project trajectory, built and launched an AI-drawing app within 2 months.• Acquired and applied expertise in various cutting-edge methods for use and deployment of large-scale image diffusion models, including controlNet, LoRAs, consistency models, and NLP prompting.

• As captain, developed detailed project roadmaps, projected team budgeting for a 2-year time period, and utilized Agile methodology to mange a team of 70+ members across 6 sub-teams to successfully initiate the first field test of team robot in collaboration with local farms.• Reached out to industry sponsors and promoted team initiatives through presenting to educational and industry audiences, improving team recruitment and obtaining over $10,000 in funding.

• As lead of navigation sub-team, led and coached a group of 8 members to develop code-base to successfully integrate software and hardware systems onboard the robot for testing of autonomous navigation.• Implemented self-driving using PID control with data from IMU, lidar, and depth sensors, all onboard an Nvidia Jetson microcontroller board.• Built custom robotic simulations using Gazebo with AWS RoboMaker to test controller and algorithms.

• Implemented a state-of-the-art image segmentation framework using PyTorch based on research papers to detect fraudulent features in NFT images.• Deployed self-supervised classification models on AWS instances and fine-tuned models on a dataset of 8 million images, boosting training speed and accuracy.• Performed statistical analysis of the latent embedding space of self-supervised classification models to optimize hyperparameters and visualized results using Weights & Biases dashboards, leading to improved model validation accuracy.

Vancouver, British Columbia, Canada

• Designed simulations for beamline physics in order to train state-of-the-art policy gradient reinforcement learning models. Successfully deployed and tested the first ever AI-controlled beamline tuning interface on the TRIUMF particle accelerators.• Wrote and published paper Accelerator Tuning with Deep Reinforcement Learning and gave poster presentation at NeurIPS 2021 Workshop.• Utilized a combination of object-oriented programming, Python multi-threading, custom shell scripting, and multi-GPU optimization to boost training time and improve usability of deep reinforcement learning architecture.• Built user interface in Python and Tkinter to communicate with the Experimental Physics and Industrial Control System (EPICS) for real-time beamline data-transfer and tuning.Reinforcement Learning” published at the NeruIPS 2021 Workshop for Machine Learning and the Physical Sciences.

Vancouver, British Columbia, Canada

• Oversaw a group of 60 first year students in a physics laboratory, coaching students on the proper use of lab equipment and data analysis techniques.• Prepared tutorials to explain the theory of experiments.• Graded and provided feedback on student’s lab reports on a weekly basis.

Vancouver, Canada Area

• Improved data-preprocessing speeds for image datasets by many orders of magnitude through designing custom scripts in Python and Bash.• Boosted team productivity by configuring custom environments in Docker to allow for research models to be trained through Huawei cloud GPU APIs.• Benchmarked and finetuned a myriad of deep-learning models for image classification and object detection in TensorFlow and PyTorch. Documented and presented findings to the team, leading to improvements on model accuracy.

University Of British Columbia

• Designed a PCB for a reflow-solder oven that interfaces with an STM32 micro-controller, utilizing a thermistor to regulate temperature and an LCD display, buzzer, and pushbuttons for user interface.• Re-designed the Brake-Light Module through determining the required circuit components to meet current demand, and improved board layout for better thermal conductivity.• Created schematics as well as 2D and 3D models of PCBs using Altium Designer for fabrication by a 3rd party contractor.

Master'S Degree, Artificial Intelligence

Pusing a masters degree in computer science with a specialization in Artificial Intelligence. I am also pursing a dissertation project.

Bachelor Of Applied Science - Basc, Engineering Physics, 86%

Activities and Societies: UBC AgroBot Engineering Design Team, UBC Formula Electric Engineering Design Team, UBC CSA student association.

Minor, Commerce

High School Diploma, 98%