MIT Jameel Clinic for AI & Health

✨🫀Thanks to genAI, we can create realistic, digital organs for virtual clinical trials. But looking realistic is easy — behaving realistically remains a challenge. A new paper accepted to @iclr_conf from @aihealthmit researchers proposes a solution: openreview.net/pdf?id=JyboUMe…

🚀MIT Flow Matching and Diffusion Lecture 2026 Released (diffusion.csail.mit.edu)! We just released our new MIT 2026 course on flow matching and diffusion models! We teach the full stack of modern AI image, video, protein generators - theory and practice. We include: 📺 Videos: Step-by-step derivations. 📝 Notes: Mathematically self-contained lecture notes 💻 Coding: Hands-on exercises for every component We fully improved last years’ iteration and added new topics: latent spaces, diffusion transformers, building language models with discrete diffusion models. Everything is available here: diffusion.csail.mit.edu A huge thanks to Tommi Jaakkola for his support in making this class possible and Ashay Athalye (MIT SOUL) for the incredible production! Was fun to do this with @RShprints! #MachineLearning #GenerativeAI #MIT #DiffusionModels #AI

🚨 High-profile breaches, such as Stanford Hospital’s $20M lawsuit over leaked records, underscore the need to audit AI models in healthcare. @sana_tonekaboni elaborates on how we can ensure electronic health records don't get exposed.

Researchers @AIHealthMIT & @Meta were recently selected to give an oral presentation @iclr_conf on GLASS Flows, a method for improving the computational efficiency of flow matching & diffusion models w/o additional fine-tuning. openreview.net/pdf?id=vH7OAPZ…

What does an @mit Professor of Electrical Engineering & Computer Science who's also a cardiologist think of scenes depicting AI and cardiac events from The Pitt? Head to our Instagram to find out! instagram.com/aihealthmit/

⚖️ Most ML models don’t make decisions — they output probabilities. Those probabilities get turned into yes/no actions using a threshold. A new paper accepted to @aistats_conf argues we should evaluate models across many possible thresholds, not just one. openreview.net/forum?id=Ot0KB…

We’ve catalogued ~250M protein sequences, but <0.1% are experimentally validated. A new paper in @NatureCatalysis introduces EnzymeCAGE: a geometric foundation model trained on 1.5M enzyme–reaction pairs to directly model enzyme–reaction compatibility. nature.com/articles/s4192…

“We may have a crisis on our hands,” says @AIHealthMIT PI @RosalindPicard on the increasing number of people sharing sensitive personal information with AI chatbots. Read more in @TIME time.com/7379564/ai-emo…

⏰ It's about time to adopt AI into routine mammograms, according to @EricTopol's latest Ground Truths newsletter, which highlights Mirai as an example of an extensively validated AI tool that can assist w/ prevention and early detection of breast cancer. erictopol.substack.com/p/why-all-mamm…

Proud to announce that clinical trials for @ARPA_H ADAPT Program have begun this week, using AI models developed by @aihealthmit and @MassGenBrigham researchers to personalize treatments for cancer patients. jclinic.mit.edu/clinical-trial…

🏅 Please join us in congratulating #JameelClinic PIs @MurielMedard and @RosalindPicard on their recognition in the 2026 @IEEEAwards! The IEEE medals are the organization's highest honors. 📰Read more: jclinic.mit.edu/2026-ieee-awar…

🦠 Can AI rise up to address the Antimicrobial Resistance Crisis & other infectious diseases? @AIHealthMIT faculty lead Jim Collins recently spoke on how his collaborations w/ @BarzilayRegina & Tommi Jaakkola have informed his work. news.mit.edu/2026/3-questio…

In a recent @NatureComms paper, @aihealthmit Fellow Evan Collins, along w/ a team of researchers across MIT, MolCube, KIST, & @LehighU propose a Lipid Nanoparticle Database, the first integrated, open resource of its kind for LNP research. nature.com/articles/s4146…

💭 If you can see yourself reflected in others, you begin to believe that their future could be yours too. @SmithsonianScie published a free ebook spotlighting 12 women shaping AI, including #JameelClinic faculty lead @BarzilayRegina! ssec.si.edu/women-in-artif…

Is it safe to share your medical info with LLMs? @HiawathaBray interviewed #JameelClinic PI @MarzyehGhassemi & @MonicaNAgrawal (who received her PhD '23 from #JameelClinic PI @david_sontag) about who is at risk when using LLMs for health guidance. bostonglobe.com/2026/02/03/bus…

ML models don't always generalize well to underrepresented patient subgroups. A new paper in @JCOCCI_ASCO proposes balancing the dataset instead of modifying the model architecture, demonstrating improved performance for patient survival. ascopubs.org/doi/full/10.12…

Tonight at the @boltz_bio meet-up in London, co-founder @GabriCorso addresses the audience. Boltz was founded at the MIT Jameel Clinic (@AIHealthMIT).

A recent paper Spotlighted @NeurIPSConf questions a comforting assumption in OOD generalization: models that do well ID also do well OOD. "In a new setting this ‘best model’ could be the worst model for 6 to 75 percent of the new data." — @MarzyehGhassemi news.mit.edu/2026/why-its-c…

🏥Can patient-driven innovation change our approach to medtech? Eye for AI host and longtime @nytimes correspondent @craigss writes about how "patient innovators" like #JameelClinic faculty lead @BarzilayRegina are transforming healthcare technology. forbes.com/sites/craigsmi…

Congrats to the @boltz_bio team on the Boston launch of Boltz PBC yesterday @MIT w/ @GabriCorso sharing what's next for Boltz and @jeremyWohlwend doing a live demo! Excited to see what's next and how Boltz will accelerate the discovery and design of novel therapeutics!