Work Experience
Suncorp
Apr 2025 – Aug 2025
Melbourne, Australia
AI/ML Engineer
OverviewLed end-to-end development of an automated LLM-prompt optimisation framework ensuring consistency and minimal LLM-drift across time and model upgrades. Collaborated with Data Science team on LLM-based solutions for motor vehicle claims processing.
ApproachImplemented an iterative optimisation loop using LLM as a judge to provide actionable feedback, applying directional prompt updates until convergence on an acceptable baseline score or fixed epochs.
OutcomesDelivered a scalable LLM prompt optimisation framework with at least 20% accuracy increase across projects, planned for rollout across all LLM-powered initiatives.
Swinburne University of Technology
Dec 2024 – Apr 2025
Melbourne, Australia
Applied ML Researcher
Project: On Forecasting, Deep Learning, and Kolmogorov-Arnold Networks (KANs)
OverviewExtended final-year thesis on stochastic data modelling with KANs, building on comparisons of ARIMA, LSTM, and KANs. Explored autoencoders for dimensionality reduction and KAN-based models for non-Euclidean data.
ApproachDeveloped statistical models, LSTMs, and KANs on DataBricks. Designed AI pipelines for prototyping and validation against MSE/RMSE. Explored temporal and textual data modelling using fine-tuned Mistral via Hugging Face.
OutcomesEstablished a baseline case study for stochastic data modelling with KANs, motivating specialised KAN-based models for interpretable forecasting.
Swinburne University of Technology
Dec 2023 – Feb 2024
Melbourne, Australia
Statistical Modelling Researcher
Project: Investigating Alternative Selection Criteria for ARIMA Models
OverviewInvestigated alternative selection criteria for ARIMA models, comparing the Hyndman-Khandakar algorithm with a new approach using Shapiro-Wilk, Ljung-Box, and t-tests.
ApproachUsed custom R scripts for diverse time-series dataset generation, applied parameter refitting to minimise overfitting, benchmarked with rolling window cross-validation on Posit Workbench and cloud HPC.
OutcomesAchieved 15% RMSE reduction vs traditional methods and eliminated 40% more insignificant parameters.
Swinburne University of Technology
Dec 2022 – Feb 2023
Melbourne, Australia
Scientific Computing Research Assistant
Project: On Efficient Calculation of Virial Coefficients
OverviewDeveloped an algorithm for computing virial coefficients in hard particle systems, optimising efficiency by identifying unique integrals among permutations.
ApproachApplied graph theory and adjacency matrices to define integration limits. Implemented distributed computations on DataBricks for large-scale integral evaluation.
OutcomesReduced computation time by 81% by identifying unique permutations among over 5000 integrals.