Ritwik Gupta 🇺🇦
@Ritwik_G
Ph.D. Candidate @berkeley_ai | Incoming Assistant Professor @umdcs | Technical Director @DIU_x
Ever wondered if the way we feed image patches to vision models is the best way? The standard row-by-row scan isn't always optimal! Modern long-sequence transformers can be surprisingly sensitive to patch order. We developed REOrder to find better, task-specific patch sequences.
There’s no other satellite company hustling as hard as Umbra. Amazing to see domestic manufacturing capacity spinning up!
We’ve scaled up! Umbra’s new 50,000 sq ft facility in SoCal quadruples our production capacity, strengthening U.S. space and defense readiness with trusted, American-made SAR. Read more: umbra.space/blog/umbra-exp…
📢 @umdcs is welcoming 8 new faculty in 2025–26 with expertise in AI policy, robotics, bioinformatics, vision, language models, simulation & sound design. They’ll strengthen research & teaching across computing. Read more: go.umd.edu/New-Faculty7-2…
Some problems can’t be rushed—they can only be done step by step, no matter how many people or processors you throw at them. We’ve scaled AI by making everything bigger and more parallel: Our models are parallel. Our scaling is parallel. Our GPUs are parallel. But what if the…
I'm at ICML w/ my co-author Jihoon this week presenting our work on using multiple visual encoders for video understanding! 💡 Thursday AM Poster: West Exhibition Hall B2-B3 #W-300 Reach out to chat about video and the holy grail of multiple visual tasks in one! #ICML2025
Have you ever wondered why we don’t use multiple visual encoders for VideoLLMs? We thought the same! Excited to announce our latest work MERV, on using Multiple Encoders for Representing Videos in VideoLLMs, outperforming prior works with the same data. 🧵
Welcome @ghadfield, who joins @JohnsHopkins as the Bloomberg Distinguished Professor of AI Alignment and Governance. An esteemed scholar, Hadfield is joining the hub for Promoting and Governing Technological Advances. hub.jhu.edu/2025/06/26/gil…
When Jane Darby talks about Iran, I listen. Well outside the normal AI stuff I talk about, but Jane Darby was an incredible colleague on campus and exposed me to new ideas around nuclear stability and JCPOA constantly!
Options for dealing with Iran's nuclear program have long been framed as binary: diplomacy/war (or at least military intervention)... but in practice, option 2 generally circles back to option 1 eventually (otherwise you get stuck in endless crises)
We’ve scaled up! Umbra’s new 50,000 sq ft facility in SoCal quadruples our production capacity, strengthening U.S. space and defense readiness with trusted, American-made SAR. Read more: umbra.space/blog/umbra-exp…
I was recognized as an outstanding reviewer for @CVPR! I made it a goal this year to think about better ways to write reviews, so I’m happy to see that effort worked.
Can we systematically generalize AR "word models" into "world models”? Our CVPR 2025 paper introduces a unified, general framework designed to model real-world, multi-agent interactions by disentangling task-specific modeling from behavior prediction.
Four years ago, before I was at Berkeley, @Ken_Goldberg gave away an AUTOLab t-shirt to the person who could best identify the tool in an image he tweeted. I won but didn't claim the prize despite sitting 5 ft away from AUTOLab. Now finally, one Ph.D. later, thanks Ken!!
I'm excited to share that I’ll be joining @UofMaryland as an Assistant Professor in Computer Science, where I’ll be launching the Resilient AI and Grounded Sensing Lab. The RAGS Lab will build AI that works in chaotic environments. If you would like to partner, please DM me!
Our work has been accepted to #ACL2025 ! Check out our paper: arxiv.org/abs/2503.04722.
Can LLM flip a biased coin? No! Can LLM update their priors with In-Context Learning? Yes! Check out our work "Enough Coin Flips Can Make LLMs Act Bayesian"