Manya Ghobadi
Associate Professor
Builds efficient computer networks and systems for machine learning and cloud infrastructure, such as network-based model checkpointing.
Manya Ghobadi's lab at MIT works on distributed, cloud & edge systems and efficient training & inference. 14 papers since 2023, 3 at top venues, mostly MICRO and PLDI. 10 PhD students. 15 alumni, most often to Apple and Jane Street. Applications close .
At a glance
- papers, 2023–2026
- 4/ year
- at T1 venues
- 21%
- PhD students
- 10
- MS students
- 0
- alumni listed
- 15
Papers by yearfrom the lab homepage, 2023–2026
- 2 papers2026 · T1 0 · collecting
- 2 papers2025 · T1 1
- 6 papers2024 · T1 1
- 4 papers2023 · T1 1
T1 venuesOtherVenues: MICRO, PLDI, SIGCOMM
Research focuspaper titles that use each topic's words
- Efficient training & inference5 papers
- Distributed, cloud & edge systems3 papers
- builds efficient0 papers
- computer networks0 papers
- infrastructure0 papers
Paper list
From the lab homepage.
20262 papers
- USENIX NSDICheckmate: Zero Overhead Model Checkpointing via Network Gradient Replication
- OFCReconfiguration-Aware Direct-Connect AI Cluster using Spatial-and-Wavelength-Selective Switching
20252 papers
- IFIP NetworkingFORESIGHT: Joint Time and Space Scheduling for Efficient Distributed ML Training (Best Paper Award)
- MICRO (T1)Spine-Free Networks for Large Language Models
20246 papers
- USENIX NSDICassini: Network-Aware Job Scheduling in Machine Learning Clusters
- ACM HotNetsMLTCP: A Distributed Technique to Achieve Optimal Flow Scheduling For Machine Learning
- PLDI (T1)NetBlocks: Staging Layouts for High-Performance Custom Host Network Stacks
- IEEE CloudNetNona: Stochastic Congestion-Aware Job Scheduling for Real-Time Inference Queries
- IEEE Hot InterconnectsRail-only: A Low-Cost High-Performance Network for Training LLMs with Trillion Parameters
- ACM HotNetsThe Case for Decentralized Fallback Networks
20234 papers
- IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) [paper][slides]InfoShape: Task-Based Neural Data Shaping Via Mutual Information
- SIGCOMM (T1)Lightning: A Reconfigurable Photonic-Electronic SmartNIC for Fast and Energy-Efficient Inference
- ACM HotNetsOn-Fiber Photonic Computing
- USENIX NSDITopoOpt: Optimizing the Network Topology for Distributed DNN Training
Where alumni went15 listed, 75% to industry
Industry 3Academia 1
Most often to
- Apple
- Jane Street
- NVIDIA
Also to
- UC Berkeley
Apply to MITFall 2027
MIT EECS Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- See program page
- English
- TOEFL/IELTS
- Fee
- USD 90
Confirmed on the program page.
Similar labsnearest by research summary
- GBEdinburgh
Michael O'Boyle
- Compilers & programming languages
- USUSC
Viktor K. Prasanna
- Accelerators, chips & architecture
- USMIT
Tim Kraska
- Large language models
- HKCUHK
James Cheng
- Distributed, cloud & edge systems
- USUMD
Abhinav Bhatele
- Efficient training & inference
- USUSC
Konstantinos Psounis
- Efficient training & inference
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
MITcloses