Latest Interviews
Showing 1–2 of 2 transcripts.
Clear all filters- Jane Street1h 6m
The Uncertain Art of Accelerating ML Models with Sylvain Gugger
Sylvain Gugger, Ron Minsky, Jeremy Howard, Mark Mandelmann, Mark Mirchandani, Francesc Campoy, Gabriel Sanchez
Former fast.ai co-author Jeremy Howard discusses his transition from mathematics education to optimizing machine learning infrastructure at Jane Street, highlighting breakthroughs in learning rate schedules and image resizing that previously secured top benchmark placements. He details the development of the Hugging Face Accelerate library, a lightweight tool designed to abstract complex hardware parallelism and eliminate boilerplate code for training across diverse GPUs and TPUs. The discussion further explores the architectural constraints of financial data, the dominance of PyTorch's iterative execution model, and Jane Street's rigorous approach to reproducibility and custom model development for high-frequency trading.
- Jane Street45 min
Clock Synchronization with Chris Perl
Responding to MIFID 2 regulations, Jane Street engineered a hybrid time synchronization system that combines GPS-driven Precision Time Protocol with Linux-based NTP Interleaved Mode to achieve sub-100-microsecond accuracy across its infrastructure. By leveraging hardware timestamping at the network interface and cross-verifying local and distant time sources, the architecture mitigates standard PTP vulnerabilities while maintaining a simple, inspectable network topology. This design successfully reduces worst-case synchronization errors to approximately 35 microseconds, effectively balancing regulatory compliance with robust fault tolerance.