Leveraging the nature of connected data with language models and graph neural networks.Developed and deployed multiple end-to-end machine learning products from ideation to production, leveraging knowledge graphs and large language models on text data. Focused on building reproducible, scalable, and high-performing recommender systems, including a system for a leading scientific publisher that achieved a 90% user preference rate and reduced outreach campaign time from weeks to hours. Engaged in applied research to tackle complex problems and drive innovation, staying current with ML advancements and integrating new techniques into production-ready products. Additionally, implemented various ML infrastructure tools to streamline the ML product pipeline, including experiment tracking, model preparation tools, MLOps platforms, and dynamic scaling solutions.

At IBM Research I worked on my Master's thesis. I developed GrafAE, a knowledge graph completion framework for multi-relational knowledge graphs. This framework integrates five knowledge graph embedding models and two autoencoder models, which can be combined to create eight additional novel models. Performance, measured by mean reciprocal rank (MRR) and hits at k (Hits@K), showed significant improvements. This work earned me the highest grade.

Contributed to multiple cloud-based solutions utilizing microservice architecture and full-stack development skills. Key projects include developing an orchestration service that used OCR to digitize drug prescriptions, significantly reducing delivery times for continuous glucose monitors. Additionally, worked on automating meeting minutes tracking, task identification, and assignment through speech recognition and natural language processing, as well as contributing to a real-time speech-to-speech translation service for seamless cross-language communication.

I taught the C++ programming language to first year math students at ETHZ.