President of Imperial College Radio Society.- Nominated for Best Society and Most Improved (of 428)- Premiered guests speakers for live in-person talk shows, including The Economist and world-leading academics. - Organised 10 sell-out music events throughout the year, with cumulative 8K+ attendees, and record ticket sales, selling out in less than 3 mins. at times.- Grew a strong and devoted community, seeing 2x membership and 3x increase in active volunteers, from last year.

- Led a tech transformation, upgrading and optimising a large tech stack from legacy systems. Crafted a bespoke website, applets, and APIs for custom studio hardware. Reduced dedicated servers by 50% and FTEs by 3.- Agile delivery approach resulting in a new portal platform going live in record 8 weeks time. Allowing scalable and automated platform that is future proof and supports composable IT architectures.

Extended work of Poggio et al. to derive new bounds for neural network complexities in discrete function spaces, previously established only for continuous functions, enabling applications and analysis in real-world scenarios such as financial time-series data.Executed large-scale experiments on High-Performance Computing clusters, running over 1,500 computational hours, to validate novel self-derived theoretical results, extensively utilising Python, TensorFlow and PyTorch. Formulated innovative methodologies for crafting highly efficient neural network architectures ,achieving significant reductions in model size—by orders of magnitude—compared to traditional shallow and deep networks, while maintaining or improving accuracy.

Focus in Machine Learning, AI, Statistics, and Finance.Courses Taken:Mathematical Foundations of ML, Optimisation, Statistical Learning, Methods for Data Science,SDEs in Financial Modelling, Intro to Option Pricing, Game Theory, Dynamics of Learning and Iterated Games

Designed and implemented advanced Reinforcement Learning Models for hedging derivatives under Market Frictions, based on previous work of Gonon et. al, seeing up to a 40% decrease in model size while maintaining accuracy compared to traditional models.Created models to approximate the super-hedging price and strategy up to maturity of a portfolio of derivatives.