Arianna P

Owned the LLM extraction layer converting raw purchase emails into structured transaction records, iterating on prompts and post-processing logic to fix currency, quantity, and deduplication errors. Built and hardened data-quality checks across a BigQuery purchases warehouse, routing affected records through a Go-based reprocessing service.

Designed and operated batch and real-time data pipelines on GCP processing hundreds of millions of events per day across 20+ enterprise clients, supporting segmentation, activation, and lifecycle personalization workflows. Integrated heterogeneous data environments into unified ingestion and transformation workflows, enabling consistent cross-client analytics on petabyte-scale event streams.

Designed and maintained real-time streaming pipelines using Apache Flink and Go on Kubernetes, backed by Apache Pulsar, to power a multi-tenant event-driven architecture for high-volume betting operations. Developed DataVault modeling in SingleStore at terabyte scale, with analytics layers orchestrated via Airflow and served through dbt+Looker and Preset across multiple business domains.

Built and maintained ETL pipelines in Airflow and AWS Glue to consolidate data from lab systems, mobile apps, and third-party APIs into structured data marts, enabling downstream analytics and personalization features. Developed a product recommendation engine using collaborative filtering on medical and behavioral attributes, serving tens of thousands of users with fully personalized nutrition an

Worked on data engineering and machine learning projects for enterprise clients in automotive and industrial sectors. Built high-volume PySpark and Scala pipelines on AWS for BMW, and developed predictive maintenance models for SKF using Python, AzureML, and Databricks.

Worked within the innovation team on computer vision initiatives to support fraud detection workflows. Built a deep learning near-duplicate image detection system using Siamese neural networks and contrastive learning, and developed an internal Django tool for image comparison and semi-supervised dataset labeling.

Activities and societies: Co-inventor of patent DRACOOLA – Driven Air Cooling Advanced System

Completed a Bachelor's degree in Computer Engineering and a Master's degree in Analytics and Machine Learning, with coursework in statistical modeling, machine learning, AI systems, and software engineering. Co-inventor of the patent DRACOOLA – Driven Air Cooling Advanced System, developed within a university research project on advanced thermal management and cooling system design.

Activities and societies: Co-inventor of patent DRACOOLA – Driven Air Cooling Advanced System

Completed a Master's degree in Analytics and Machine Learning, with coursework in statistical modeling, machine learning, AI systems, and software engineering. Co-inventor of the patent DRACOOLA – Driven Air Cooling Advanced System, developed within a university research project on advanced thermal management and cooling system design.