Audit data analyst in Dubai. I build the Python and Caseware IDEA analytics behind journal entry testing for statutory audits: the pipeline that checks a ledger is complete, tests it for the characteristics auditors look for, and turns thousands of flagged entries into a short, explained list the audit team can select from.
MSc Data Science (Distinction), University of Surrey · BSc Computer Engineering, UC San Diego
gl-risk-scoring: ranks journals for ISA 240 journal entry testing, with a plain-language reason for each, in an Excel working paper. On 30 synthetic ledgers its top-50 list is 85% precise, against 67% for counting tests and 57% for an Isolation Forest, with an ablation, robustness checks and one honest tie reported. A clean-room version of tooling I built at work.
Python (pandas, NumPy, SciPy, scikit-learn, pytest) · SQL · Caseware IDEA and IDEAScript · Power BI · Git and GitHub Actions
Audit and forensic data analytics · explainable scoring and evaluation · LLM evaluation (my MSc dissertation looked at detecting hallucinations from a model's internal signals)