Mohammad Alizadeh

I work at the intersection of computer networks, systems, and machine learning, with the aim of making large-scale computing infrastructure faster, more robust, and easier to manage.

I joined MIT in 2015. Before that I completed my Ph.D. at Stanford University and spent a few years in industry, at the datacenter networking startup Insieme Networks and at Cisco. My earlier work on datacenter transport and load balancing — DCTCP, pFabric, CONGA — is implemented in commercial operating systems and switching products and deployed by large datacenter operators. More recently my group has focused on using machine learning to design, simulate, and optimize computer systems.

My work has been recognized with the ACM Grace Murray Hopper Award (2022), the SIGCOMM Test of Time Award (2021), an Alfred P. Sloan Research Fellowship (2017), the SIGCOMM Rising Star Award (2017), an NSF CAREER Award (2018), a Microsoft Research Faculty Fellowship (2019), a VMware Systems Research Award (2019), and best paper awards at SIGCOMM (2014, 2017, 2022), NSDI (2023), and SIGMOD (2021).

In the department I serve as Industry Officer and as Director of the EECS Alliance Program.

My current work centers on bringing machine learning to bear on the design and operation of computer systems, alongside continuing work on network protocols and transport:

• AI for automated systems design. Can an AI autonomously design mechanisms for computer systems with the creativity and reasoning of a human expert? Glia uses large language models in a human-inspired, multi-agent workflow to produce interpretable designs for distributed systems problems, and exposes its reasoning.
• Learning-based performance modeling and simulation. Models that predict how a system will behave — compositional analytical-ML models of CPU performance (Concorde), flow-level network performance estimation (m3), and counterfactual trace-driven simulation (CausalSim).
• Decision-making and control with learning. Reinforcement learning for resource management, scheduling, and adaptation in non-stationary, context-driven environments.
• Systems for machine learning. Improving the performance and scalability of large-scale training and inference, including pipeline parallelism (GraphPipe) and efficient video inference on edge devices.
• Networks and transport. Congestion control and internet congestion management, datacenter transport, low-latency RPCs, and real-time video and videoconferencing systems.
• Learned database systems. Query optimization and cardinality estimation with learned models (Bao, Neo, FactorJoin), in collaboration with the MIT Data Systems Group.

I teach computer networks and systems, including 6.829: Computer Networks, 6.888: Advanced Topics in Networking, and 6.02: Introduction to EECS II — Digital Communication Systems.

I have been fortunate to work with a terrific group of Ph.D. students, post-docs, and master’s and undergraduate researchers. If you are a current MIT student interested in working with me, please send me an email and we can arrange a time to meet.

Prospective students: please apply to the EECS graduate program and mention my name in your application.