MLOps

• Responsible for our machine learning infrastructure, developing distributed and accelerated data pipelines and ML model frameworks.

• Improving our data science workflow, producing tooling and services to aid in designing, deploying and maintaining analytics.

• Implemented a new analytics dashboard, to preview analytics and evaluate performance across millions of events. The integration of an accelerated distributed runner resulted in approximately a 10 times improved in speed, and the ability to extract valuable performance metrics.

• Added the ability to dynamically generate a score for our rule based fraud detection language, allowing for model ensembling and scaling.

Deep Learning Secondment

• Entered an international competition for financial crime detection, Privacy Enhancing Technologies Challenge, hosted by Innovate UK, in partnership with the UK and US governments. Collaborated with the research team to develop and submit a privacy preserving, federated deep learning solution.

• Responsible for engineering a prototype model, building on the research code and white paper. Adapting it to meet the constraints of the competition and mandatory federated learning framework flower. Optimizing performance and containerizing, ensuring testability and scalability.

• The solution achieved the highest accuracy score in the UK competition track for its ability to detect financial crime, returning an average “Area-Under-Precision-Recall-Curve” of 0.94. The solution was also singled out with a special award for its usability and production readiness.

Summer Internship Programme

• Co-led a team of six interns to deliver a project which added metadata to block and allow lists, and enabled sorting and filtering of items through the UI.

• The first intern programme to be merged into the release product and completed on time.

• Designed and planned the feature, breaking into individual manageable sub tasks for the interns.

• Responsible for sprint planning, review and hosting daily stand-ups.

• Onboarding and introduction to our systems and software development cycle.

• Mentored, giving guidance and pair programming.

• Reported high level status updates on feature progress to the directors.