Tariq Soliman

Oversee financials, fundraising, and governance for a community sport club; contribute to growth and organisational sustainability.

Investigated dimension importance estimation with embeddings generated by Large Language Models for information retrieval, building on prior work in dimensionality reduction and retrieval efficiency.

Supported a digital twin of the Dugald River Mine (DRM) zinc flotation processing plant, fixing bugs and adding features. The digital twin uses neural networks to simulate operations and Particle Swarm Optimisation to optimise control variables for zinc grade and recovery. The project increased plant recovery by 0.27% (p < 0.01), reduced model errors in one stream by over 20%, introduced unit testing into CI/CD pipelines on Azure DevOps, and engaged with clients in weekly feedback meetings. Investigated error detection and correction techniques for the DRM Flotation Twin (e.g., ensembles, Mean Variance Estimation Networks) to estimate uncertainty and added explainability metrics (Sobol Analysis, SHAP) in training with Azure Databricks.