High throughput trade processing systems

• Worked on increasing throughput for the memory hierarchy of Berkeley’s open source RISC-V cores, Rocket and BOOM.

• Wrote baremetal tests to benchmark L2 and LLC performance during sequential reads and writes.

• Made changes to MSHRS in the L1 and L2 for more consistent performance.

• Scaled up a service used for consolidating trade executions to over 10x its previous capacity along with a significant improvement in performance.

• Created a Slackbot which allows internal clients to query for transactions with multiple search filters.

• Researched pre-filtering algorithms for a display which vision impaired individuals could view without corrective lenses.

• Developed ray tracing and light transport models to support the generation of vision corrected images.