Mallory Brickerd

Smart Matching Engine at Eraneos1 Bundesbahnen (SBB) . Implemented an internal matching engine to enhance the data quality of SAP data, addressing gaps in their records of spare parts used across various trains and their suppliers. By leveraging graph techniques and semantic matching of part and train descriptions, we successfully identified and matched 25% of previously unlinked or missing train parts. Demonstrated the engine's ability to accurately match parts to trains and identify similar suppliers, significantly improving the integrity and usability of SBB's data management system and providing a scalable foundation for ongoing data quality enhancements.

DOR API at Dusseldorp . Developed an Azure Function using FastAPI to integrate with the Digitaal Opkopers Registraar (DOR) for Dusseldorp Automotive. This solution automated the official registration process within the DOR API, ensuring compliance with newly introduced regulations surrounding car purchases. By providing near real-time logging and tracking of transactions such as vehicle sales, purchases, and inter-dealer transfers, the implementation significantly improved operational efficiency and accuracy. Enabled Dusseldorp Automotive to seamlessly align with legal requirements while enhancing their workflow and empowering dealers to monitor transactions with greater ease and confidence.

Optimal Scenario Planning Developed a demo tool in Python to simulate and optimize production processes. The tool models key production floor components such as prescription orders, employees, machines, and stock levels, allowing users to adjust parameters like production speed, machine utilization, and stock availability. By leveraging digital twin techniques, it enables the client to test multiple scenarios, identify optimal strategies, and improve resource allocation without disrupting real-world operations. This solution reduces downtime, minimizes waste, and supports data-driven decision-making for high-volume, time-sensitive production environments.

Global Procurement Optimisation at Heineken . Developed a global spend analysis solution for Heineken to optimize procurement and reduce unnecessary orders by identifying part matches across suppliers and factories worldwide. Leveraging semantic matching techniques, we provided insights into spare part usage, supplier pricing, and stock availability. This solution enabled Heineken to identify cost-saving opportunities, streamline procurement processes, and improve operational efficiency on a global scale.

R2Py at Eraneos1 . Translated the AzureML ARIMA pipeline from R to Python and PySpark to enhance maintainability and modernize the Azure Data Factory sales forecasting pipelines. This involved assessing and re-engineering R-based time series forecasting logic for thousands of products across health and hygiene departments, ensuring equivalent efficiency and parallelization while leveraging expertise in Python, PySpark, and time series forecasting to deliver a robust and scalable solution.

AWS Machine Learning Platform & Pipeline at Tinka . Implemented an automated end-to-end MLOps pipeline in AWS, leveraging the latest SageMaker capabilities to build a robust ML platform. Utilized SageMaker's advanced tools, including Data Wrangler, to prepare and transform the data. Used SageMaker model containers to train and deploy models for predicting the likelihood of repayment after selecting the "pay later" option at checkout on Wehkamp's website. This solution streamlined data preparation, model training, and deployment processes, providing Tinka with a scalable and automated platform for accurate, data-driven decision-making.

Scalable MLOps Platform at New10 . Implemented a scalable AI/ML platform and cloud infrastructure for New10 using AWS SageMaker, enabling the data team to work more efficiently and scale their solutions following industry best practices. Set up a robust MLOps pipeline to automate data ingestion, preparation, model training, and deployment, while integrating New10's existing locally developed regression models using custom SageMaker model containers. This solution established a consistent and reusable MLOps framework, empowering the data team to streamline workflows, accelerate model development, and ensure scalable deployment of AI solutions. By modernizing the infrastructure, we improved collaboration, reduced time-to-market, and provided New10 with a strong foundation to optimize decision-making and drive business growth.

Tosho Optimisation at Apotheek Voorzorg . Built a digital twin of Voorzorg's production process by deeply analyzing and restructuring historic production data, enabling offline simulations to optimize machine configurations and pill allocations without disrupting operations. Through these simulations, we identified a 2-week optimization period as optimal, aligning with the recurring prescription cycle where patients reorder medications. Using a custom heuristic algorithm, we provided actionable recommendations to maximize automated prescription fulfillment. This solution delivered significant operational improvements and achieved €500K in cost savings within the first year.

User-Based Recommend Engine at Pathé . Built and deployed a user-based recommendation engine using matrix factorization techniques to personalize movie suggestions for e-mail marketing campaigns. The solution enabled the e-commerce team to deliver targeted recommendations, enhancing customer engagement and driving higher conversion rates. I worked closely with an engineer to integrate the recommendation engine into the existing deployment pipeline, utilizing Airflow for automated workflow orchestration. This project successfully introduced data-driven personalization to email campaigns, creating more tailored and relevant experiences for customers.