Vita

Last updated: 09/09/2024

Summary: Looking for engineering roles in AI. I have a strong foundation in mathematics coupled with advanced programming ability. This enables me to quickly iterate on and implement complex ideas. I am naturally optimistic, an effective team player, a quick learner, and possess high stamina.

Here is some illustrative work, see my GitHub and website for more.

1. Eval Framework: Developed LLM testing infrastructure for sea.dev. The evals test the ability of LLMs to extract information from multi-turn dialogue, implementing recursive prompt architectures to measure information extraction accuracy across conversation depths.
2. Sherlock Holmes Eval (video/repo): Test a LLM's ability to identify the culprit in murder mysteries.
3. Guaranteeing JSON from GPT-2 (video/repo): Reproduced GPT-2 on 8xA100 GPUs, fixed a subset of tokens in the output sequence to be JSON schema leaving gaps for model to generate the values.
4. Training algorithm for regularized models on arbitrarily large data sets (video/arXiv/repo): Eliminates requirement of storing O(n) quantities in memory, and allows for training data to be stored in distinct sites for privacy concerns.
5. Debiased high-dimensional logistic regression (video/arXiv/repo): Conjecture to debias model in setting where both the number of features and observations are asymptotically increasing.

Technical Skills

• Languages: Python, R, JavaScript, SQL
• Packages, Frameworks: PyTorch, CUDA, HPC clusters, React, Django, AWS
• Expertise: numerical optimization, computational linear algebra, and ML theory (transformers, stochastic gradient descent, deep learning, Markov processes)

Education

PhD in Mathematical Statistics. University of Warwick. Oct 2020 - Aug 2024.
Supervisor: Prof. Ioannis Kosmidis
Publications: On fitting models to very large data sets and debiasing high-dimensional logistic regression, see my Google Scholar.

MSc in Statistics. University of Warwick. Oct 2019 - Sep 2020.
Dissertation: Bias-reduced logistic regression in high dimensions and model estimation with arbitrarily large n.
Awards: Winton Award (Finished 1st in class)

BCom Actuarial Science with Honours in Statistics. University of Cape Town. Feb 2015 - Nov 2018.
Awards: STA4006W Class Medal (Finished 1st in class), Commerce Faculty Scholarship, Dean's Merit List.

Experience

LLM Engineer. sea.dev. Nov 2024.
Built multi-turn dialogue evaluation framework measuring LLM extraction accuracy.

Data Scientist. Barrows Global. May 2019 - Aug 2019.
Cleaned and extracted features from major retailer's data, built models to analyse and draw inference on the sales performance in the store environment to inform the advertising strategy for clients.

Actuarial Intern. Old Mutual Limited. Feb 2017.
Performed Value at Risk (VaR) modelling and worked with the reinsurance team to rectify a database error which produced duplicate policy holders.