Darius Grassi

About

A picture of me

I am a computer science graduate student at the University of Texas at Austin specializing in networked systems and ML-driven optimization.

Prior to UT, I was an undergraduate at the University of Illinois Chicago, where I worked with Brent Stephens. During my undergrad, I also spent a few months at the Open Networking Foundation (now Intel NEX).

Projects

ML-Driven Memory System Auto-Tuning University of Texas at Austin, 2024-2025

I designed an automated tuning system for zswap compression parameters using PyTorch, employing feature engineering to optimize performance. To evaluate the system, I built a Redis benchmarking pipeline to identify optimal system configurations for runtime. The results demonstrated that the online learned system could discover the optimal configuration in a fraction of the time required by traditional search methods.

Programmable NIC Memory Management University of Texas at Austin, 2023-2024

I prototyped a memory management system for programmable NICs to support cloud applications. This involved implementing various allocation policies using DPDK, a kernel-bypass framework essential for achieving high-throughput packet processing in modern data centers.

FPGA Acceleration Framework University of Texas at Austin, 2023

I developed a High-Level Synthesis (HLS) implementation of a CNN accelerator, which involved a detailed analysis of its performance and resource utilization. Additionally, I designed optical flow processing accelerators using HLS for an FPGA accelerator streaming platform.

Software Network Switch with RDMA University of Texas at Austin, 2022

I built a custom software switch using DPDK for kernel-bypass processing on a 100 Gbps testbed. A key part of this project was integrating RDMA for high-speed communication and conducting a thorough evaluation of TCP and RDMA performance characteristics to understand their trade-offs.

Programmable Network Testing Infrastructure Open Networking Foundation, 2021

I built a CI/CD pipeline for programmable switch regression testing using Python and Jenkins, which automated the testing process. I also designed and implemented an automated QA infrastructure for 5G UPF testing, a critical step that prevented production slowdowns. To showcase our work, I developed line-rate network monitoring demos using Grafana for industry presentations and contributed to open-source codebases through GitHub and Gerrit code reviews.

High-Performance Network Switch Architecture University of Illinois at Chicago, 2019 – 2022

I developed traffic generation tools and a benchmarking infrastructure to evaluate programmable switch designs. This work included deploying custom architectures on hardware and using realistic applications to measure their throughput and performance under real-world conditions.