HomeML engineer building research, training and inference infrastructure in Python, JAX and Rust.
Designed and implemented an event-based live trading system in a team of 12, working closely with the execution team in Zug. Led the design discussions to make our combined deployment highly available and resilient.
Built and owned the ML research platform for MFT and HFT, including high-throughput Rust services for data ingestion into ClickHouse and a centralized warehouse with over 1 PB of market data. Reduced Rust CI build runtime by about 10x.
Built and ran an LLM inference stack for open-weight models on spare GPU nodes, resilient to evictions, serving over 40B tokens per day to internal agent tooling. Evaluated open-weight models as a replacement for commercial APIs in token-heavy systems such as agent memory.
Drove the move from MATLAB bilinear models on a laptop to ML in Python, starting with deep vision models, across five teams, 50 people total. Pushed it bottom-up and onboarded other YieldStar products onto the platform.
Increased model training runs 100x by introducing Azure Kubernetes clusters, with Coder workspaces for researchers and developers. Brought the inference stack onto the YieldStar machine itself: k3s inside a Linux VM on Windows.
Delivered an end-to-end customer application for training, evaluation and inference of ML models across cloud, customer infrastructure and edge devices. Received numerous commendations for collaboration and mentoring of colleagues.
Led the design and industrialization of ML algorithms for next-gen YieldStar tools.
Analyzed optical signal data, developed model architecture and training from research to production. Implemented and optimized the end-to-end training pipeline to fit on constrained customer hardware.
Supervisor: Henk T. C. Stoof. Thesis: Excitons in 2D Semiconductors and Topological Insulators.
Supervisor: Henk T. C. Stoof. Thesis: Holographic Massive Fermions.
Finished Cum Laude (GPA: 8.5). Results peer-reviewed and published in April, 2018.
Supervisor: Yolanda Lozano. Thesis: Introduction to String Theory.
Premio Fin de Grado (Summa Cum Laude, highest grade).
Open-source contributions. Have maintained a personal cross-cloud Kubernetes cluster since 2022.
Avid Arch Linux and Neovim user.
Drives continuous personal improvement, not by doing more but by doing it better.
My family, our two cats, reading books, playing video games and weightlifting motivate me to keep learning and growing as a person.