United Kingdom
- Understood the operations of the Merchant Management and Risk departments, and their database. - Applied this understanding to design key metrics such as churn prediction and create new dashboards to visualize insights that led to improving and optimizing their targeted voucher campaigns. - Built an ETL pipeline from the staging and production database to create aggregated view tables for their new credit risk model, along with engineering additional features useful to the model.
- Contributed to ongoing research on improving the scalability and performance of Muzero, a deep reinforcement learning algorithm, using the Mindspore framework and distributed machine learning. - Used profiling techniques to identify the bottlenecks and inefficiencies of the Muzero algorithm. - Fine tuned model hyperparameters to minimize training convergence time and prevent memory issues. - Assisted in the development of the Monte-Carlo Tree Simulation (MCTS) module. - Maintained codebase including testing and debugging to ensure correctness of code implementation.