Data Science Lead at Auror in Auckland, NZ, leading a team of ML engineers and data scientists delivering AI-powered loss prevention for 100+ retailers across the US, UK, Australia, New Zealand, Chile, and Mexico, helping prevent millions in retail crime annually and making communities safer. Curating the AI, Data & Ethics track at NZ Tech Rally 2026 and developing board governance skills through the OB Program. M.Sc. Electrical Engineering, UTFPR, Brazil (2018–2020), with focus on systems optimisation and ML.

• Team: 1 Principal Data Scientist · 2 Machine Learning Engineers @ Auror
• Impact: AI & Data Science systems deployed across 100+ retailers in 6 countries (US, UK, AU, NZ, CH, MX), preventing millions in retail crime annually
• Track Lead: AI, Data & Ethics at NZ Tech Rally 2026
• Board Observer: OB Program 2026, Auckland startup governance

Identifying the same individual across fragmented retail crime records

Retail crime data arrives from thousands of stores with inconsistent naming, missing fields, and duplicate entries. This pipeline resolves identities across those records to build a unified view of repeat offenders, enabling smarter prevention at the network level.

• Tech: Azure Container Apps · scikit-learn · GitHub Actions
• Scale: Millions of crime events processed across retail networks

Matching faces across retail crime incident reports

Links faces captured in different store incidents to surface repeat offenders who operate across locations, turning isolated events into a connected intelligence picture.

• Tech: Azure Face API

Assessing third-party face recognition providers for bias and fairness

Before deploying any face recognition system in production, we evaluated multiple providers against fairness benchmarks across demographic groups. The goal: ensure the technology performs equitably and doesn't amplify existing biases in policing and retail environments.

• Tech: Python · provider APIs · fairness evaluation frameworks
• Focus: Demographic parity · false positive rate analysis · responsible AI decision framework

AI-powered voice interface for submitting retail crime incident reports

Store staff can report retail crime incidents via voice reducing friction at the point of capture and improving data quality. The system transcribes, structures, and routes reports automatically using a multi-model AI pipeline.

• Tech: Azure Speech To Text API · Azure AI Foundry · BAML

Identifying ORC activity through millions of crime events across retail networks

Organized Retail Crime (ORC) involves coordinated theft across multiple stores and is significantly harder to detect than individual incidents. This system identifies ORC patterns by analysing crime event signals at scale, enabling proactive intervention before networks expand.

• Tech: Google Vertex AI · LightGBM · Snowflake · dbt · Dagster