Muhammad Shahbaz

Missed the #P4 Developer Day about SpliDT? 🔴 Watch it: youtu.be/mPU4_Kqrmdg Speaker: Murayyiam Parvez 💡 All previous Dev Days Recordings and Slides: p4.org/p4-developer-d…

🗓️ TOMORROW: March 19 at 11 am ET/4 pm CET ✨ P4 Developer Day || SpliDT: Partitioned Decision Trees for Scalable Stateful Inference at Line Rate with Murayyiam Parvez 🎟️ Register here: p4.org/event/p4-devel…

Gigaflow introduces pipeline-aware sub-traversal caching with Longest Traversal Matching (LTM), decomposing vSwitch pipeline traversals into reusable segments that maximize SmartNIC table utilization and significantly improve cache efficiency. Read more: p4.org/gigaflow-moder…

Celebrating Professor Kang G. Shin, retiring after 40+ years of pioneering work in real-time systems, embedded systems, and mentorship at @UMich. His impact on research and generations of students is immeasurable.👏💛💙 🔗cse.engin.umich.edu/stories/profes… #UMichCSE #Leadership @kgshin1020

Transport is the next frontier in accelerating foundation models, and getting there means "Reimagining RDMA Through the Lens of ML"! In our upcoming paper in IEEE CAL'25, we explore how a domain-specific focus can supercharge transport for ML workloads. arxiv.org/pdf/2510.16606

Learn how P4 is shaping the future of programmable networks. Stop by booth C47 at the OCP Global Summit and talk to P4 community members. Don’t miss the demo today “Gigaflow: Pipeline-Aware Caching in Virtual Switches with P4” from 2-3pm #P4 #P4Workshop

We are getting closer to the NINeS deadline (October 1). Watch Scott talk about why you should submit your work to NINeS!

Sample from CSE 598 - gitlab.com/mshahbaz/mshah… Sample from EECS 498 - gitlab.com/mshahbaz/mshah… @UMichCSE | #ComputerNetworks - 2/2

Loving the LECCAP system, leccap.engin.umich.edu, at @UMichCSE, for the two courses I'm teaching this fall: CSE 598 and EECS 489 (200+ students), covering everything networks and beyond! What’s impressive is how it records and later sends a concise summary of lectures. 👇 - 1/2

Sparse workloads scale poorly due to network bottlenecks. Our recent work in #MICRO2025, NetSparse, pushes communication into NICs/switches to slash traffic & boost performance by up to 38x on 128 nodes! 🚀 Huge kudos to Gerasimos & team at @UofIllinois and @UMichCSE! 🎉

The P4 Workshop CFP is now open!

Don’t miss the opportunity to present a talk, poster or demo at the 2025 P4 Workshop - the CFP is now open and will close August 1st! View CFP and submit your proposal | tinyurl.com/yvw47cfw #P4 #P4Workshop #OpenSource

The 2025 P4 Workshop will be a hybrid event taking place October 13th. The CFP is now open - check the web page for more exciting details | tinyurl.com/bdtsrxdv #P4 #P4Workshop #OpenSource

DIRECTING DATA TRAFFIC: @UMichCSE researchers developed Gigaflow—a new way to temporarily store memory—to direct heavy traffic in cloud data centers caused by AI and machine learning workloads. bit.ly/3XR3SqV

Don’t miss the next P4 Developer Days event on June 4th, “Pipeline-Aware Sub-Traversal Caching for Modern Smart NICs.” Learn more and register | tinyurl.com/ycmsfr67 #P4 #SmartNIC #OpenSource

Don’t miss a virtual P4 Developer Days event June 4th on “Pipeline-Aware Sub-Traversal Caching for Modern Smart NICs.” Learn more and register | tinyurl.com/ycmsfr67 #P4 #SmartNIC #OpenSource

📢🎉 Thrilled to share that NextGArch Lab’s projects, Gigaflow and SpliDT, have been accepted into this year’s #GSoC25 cohort! The projects will be led by Advay Singh (from @UMichCSE) and Sankalp Jha—supported by an incredible team of mentors from @Google and NextGArch Lab!

FASTER AI TRAINING: A new system, OptiReduce, reaches target accuracy faster by setting time boundaries on cloud servers during training—showing deep learning’s resilience allows approximating lost data without compromising accuracy. Read more: bit.ly/3GAkr4A

OptiReduce speeds up #AI and #machinelearning training across multiple cloud servers by setting time boundaries rather than waiting for every server to catch up, according to a study led by CSE Prof. @msbaz2013. ➡️More: cse.engin.umich.edu/stories/perfec…

As AI and machine learning workloads continue to grow, solutions like Gigaflow improve virtual switches for programmable SmartNICs, delivering a 51% higher hit rate and 90% lower misses. 🔽Learn more [article]: cse.engin.umich.edu/stories/stream…… #CloudComputing #AI @msbaz2013