Vinith Suriyakumar
MIT EECS, LIDS, IMES, Bridgewater AIA Labs, Stanford
I am a fifth year PhD student at MIT EECS where I am advised by Dr. Ashia Wilson and Dr. Marzyeh Ghassemi. I also collaborate frequently with Dr. Dylan Hadfield-Menell. I’m supported through a fellowship from Bridgewater AIA Labs led by Dr. Jas Sekhon where I also conduct research part-time.
Broadly, I’m interested in the privacy, security, and safety of machine learning. I’ve worked on many topics in these areas including differential privacy, auditing, algorithmic fairness, and unlearning.
My research focuses on securing internet platforms against malicious behavior enabled by generative AI systems. I am particularly interested in the safety, security, and governance of open-weight foundation models, with recent work focusing on image-based abuse such as CSAM and NCII generation. Additionally, I focus on addressing safety issues related to cyber misuse, investigating new pretraining algorithms to address dual-use risks. I also investigate security challenges in closed-source systems, including jailbreaking and secure fine-tuning interfaces.
My research advancing the privacy, security and safety of machine learning has received awards at NeurIPS, ICLR, ICML, and FAccT:
- ICML 2022 Spotlight
- ICML 2023 Oral
- ICLR 2024 Oral
- FAccT 2024 Best Paper Award
- NeurIPS 2025 Spotlight
news
| Dec 2, 2025 | I’ll be a Visiting Student Researcher at Stanford with Dr. Sanmi Koyejo from February to June 2026. |
|---|---|
| Oct 18, 2025 | I gave two invited talks recently on open-weight model safety and unlearning at the MITAI Conference and the Bridgewater AIA Distinguished Speaker Series. |
| Sep 18, 2025 |
Our work uncovering and formalizing spurious correlations originating from syntactic shortcuts in LLMs, Learning the Wrong Lessons: Syntatic-Domain Shortcuts in Language Models Shaib, was awarded a Spotlight at NeurIPS 2025! This work provides a new perspective on different forms of memorization and uncovers a new type of jailbreak in LLMs. |