Francesco Dondi

Contributed to cutting-edge research tackling the ARC challenge. Especially focused on data augmentation. Extensively used BART for classification, and Pytorch/Lightning for efficient multi-GPU training.

Had the opportunity to have direct experience of Computer Vision, both on cloud and edge embedded devices (Nvidia Jetson platform). Contributed by overhauling Docker setup and solving bugs.

Developed infrastructure around ClickHouse SQL, deploying with Salt- Stack, Helm and Terraform. Go for Telegraf plugins.

Developed the SQL optimization engine and testing infrastructure. * Expanded supported SQL syntax and optimizations * reworked the test framework to support new test suite * contributed to rewriting to higher standards the optimization engine

* Maintained and developed a Trading Engine (C++) * redesigned, implemented and migrated storage (PostgreSQL) * extracted and integrated data from multiple log streams * analyzed data to highlight opportunities for improvement (Julia)

Within Search, productionized an initially experimental tool (C++). * Improved runtime from days to hours * added support for new kinds of data * made the results easily visualizable and monitorable * made onboarding orders of magnitude faster for most cases