I studied MSci Chemistry at Imperial College London, spending my final year at MIT — the only student from my cohort selected for the exchange. Graduated 1st class (83%, 5.0 GPA), Dean's List all four years, departmental prize, and best MSci project in physical chemistry.
Dual-server AI architecture for a portfolio management team. Design to production in two months.
Early hire at an AI-native energy storage startup. Built and shipped production digital twins for battery cells and packs.
Built a Python/Dash dashboard for lab cycling data visualisation, used across active experiments.
Gaussian Process surrogate model for Li-ion thermal runaway. Published in Computer Aided Chemical Engineering.
PV module test data analysis; industry trend graphics published on the PVEL blog.
UX and The Democratisation of AI May 2026+
Millions of white-collar workers are using AI to optimise their workflows. Yet, the gap between those who use it cleverly versus those that don't is growing.…
Open in Substack →Opinion: OpinionsMar 2026+
Why do strong opinions matter when founding an important company?
Open in Substack →Zero-Dimensional Digital Twin for Redox Flow BatteriesHigh-fidelity stack-scale RFB simulation generalised for any redox chemistry. Simulates days of cycling in seconds.Digital TwinRedox Flow BatteriesPhysics-Informed ModellingPython+
Developed during my master's at MIT (Brushett Lab). Presents a zero-dimensional model generalised for any redox chemistry and operating conditions — capable of simulating stack-scale cycling at high fidelity in seconds. Enables fast design iteration and real-time control applications. Awarded best MSci project in physical chemistry at Imperial. Published in the Journal of The Electrochemical Society (Q1).
Gaussian Process Inference of Li-ion Thermal Runaway KineticsSurrogate model that extracts failure dynamics from sparse, noisy experimental data — making physically unobservable states predictable.Gaussian ProcessesThermal RunawayLi-ion BatteriesSurrogate Modelling+
Built during a research placement at the Faraday Institution. Uses Gaussian Processes to infer electrolyte decomposition reaction networks from sparse concentration profile data. Latin Hypercube Sampling generates candidate frequency factor sets; the GP surrogate inverts this to estimate kinetic parameters from real experimental data. Robust parameter optimisation via continuous rank probability score. Published in Computer Aided Chemical Engineering.