• Increased downstream accuracy from 84% to 96% by designing application specific transformer/CNN models, training and augmentations• Proposed and implemented a robust autolabeling algorithm that resulted in a 1500% increase in dataset size and therefore parameter efficiency• Incorporated new, business-critical tasks using multitask learning while maintaining downstream accuracy and latency• Architected statistical optimization postprocessing layer for sequential outputs and well calibrated model training, to cut down the model's decision-taking delays by 250%• Improved sequence based testing and tooling, resulting in faster iterations and more representative testing. AI system validated running realtime in multiple popular venues for 1 year (5,000 monthly viewers).

• Proposed and trained the deep learning anti-bot effort at Monsta Infinite, a relatively novel way of tackling the bot problem that proved very effective.• Designed data generation techniques, transformer training loops and augmentations, innovating to achieve and exceed usable performance• Designed a tamper-proof data pipeline for logging built into the core game engine. Provable data integrity and low overhead.• Solution was 10 times faster and 12% more accurate in live AWS production than the initial 'mixed-integer programming' detector. Served 10,000 active users

• Appointed after graduation to lead development of Nottingham's new virtual labs effort. (Unity game engine)• Recreated complex biomedical lab experiments in its entirety in Virtual Reality.• Very short demo video here: youtube.com/watch?v=bVX-6ip9bBA

Working as an RA to continue developing the “Virtual Reality in Building Information Modelling” research project. The project involves building an online environment where civil engineers can interact with the BIM information and each other (online multiplayer). The project started under the dean’s undergraduate scholarship in 2018.Developed on Unity and Oculus.

• Picked up technologies quickly for an analytics web app and its custom companion embedded device• Dedication and success led me to secure further contract positions, as well as part-time work