Scientist @pmcc_ai @UHN PhD @sickkids @UofT @VectorInst AI for biomedicine 🤖🧬 art, fashion, design & culture ✨ don't mistake my kindness for weakness 💖🌶
Toronto, ON
Joined September 2019
- Tweets438
- Following439
- Followers294
- Likes4.2K
Valli Subasri retweeted
Language models can learn new tasks from just a few examples in their prompt, with no retraining. This is called in-context learning (ICL). Now genome language models can do it too.
With a handful of examples, Evo 2 ICL generates proteins, de novo antibodies, CRISPR-Cas9 loci, mitochondrial genomes and entire bacteriophages.
We can even use the synthetic sequences iteratively to construct a chain-of-thought (CoT) and achieve harder design goals.
New work with Yukun A. Hao, @BrianHie and team 🧵
Valli Subasri retweeted
We looked into this a bit in our benchmark of human and AI reviews (biorxiv.org/content/10.64898…)- found that they were quite complementary. Humans focus on a work's contribution while AI reviewers focus on technical verification.
Interesting work! Although I think the bigger question isn't whether we should leverage AI in the review process, but rather how best to integrate it, including at which stages, in what sequence, and how it should complement human reviewers, etc.
Lots of debate this week on whether journals should ban AI in peer review.
We actually tested this in our new preprint, and the results were pretty clear to me.
For context, I've reviewed for most of the big journals and evaluated 1,000 human reviews as an editor. I love to read and evaluate science.
TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵
Valli Subasri retweeted
"AI safety needs to be addressed at the level of the human–AI system, not the AI model alone," writes Pietro Michelucci in a new #ScienceEditorial. scim.ag/4zSzwEV
ALT A quote in bold text reads: "... a kill switch cannot by itself serve as a viable or sustainable safety feature." Attributed to Pietro Michelucci from the Human Computation Institute.
Valli Subasri retweeted
Things are moving so fast that everyone seems to have completely missed how insane it is that AI companies have been allowed to unleash agentic AI products with no regulatory guidelines.
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
Valli Subasri retweeted
Today, we introduce Priming Models, a new paradigm for training one-step generative models from scratch.
No diffusion teacher. No flow map. No CFG. No JVP.
We believe this work will challenge how we think about building one-step models.
Joint work with @sitanch, @du_yilun, @parkwbin03, Bohyung Han.
🔥 1.19 FID one-step on ImageNet 256×256 while using substantially less compute, e.g., 12X less than Drifting-B, 6X less than Drifting-L
🎨 Scales to text-to-image generation (GenEval 0.82)
🤖 World modeling on RT-1 / Bridge-V2
Priming Model is built on one core message: [Reconstruction pretraining --> Distributional post-training].
Valli Subasri retweeted
AI Costs? Sometimes the best clinical decision is to do less: Important editorial @NEJM_AI about how AI plays into costs ai.nejm.org/stoken/default+d… Relevant to the $1billion in add'l costs reported by @BCBSAssociation CC @chrissyfarr @aafp
Valli Subasri retweeted
Among US #clinicians, use of an #AI-enabled clinical information platform increased substantially over 2.5 years, with sustained use across #physician and nonphysician groups.
Queries most often focused on #treatment and #management rather than #diagnosis, indicating that these tools are being used primarily to support #clinical decision-making at the point of care.
ja.ma/4Ba9ej2
Valli Subasri retweeted
AlphaGenome made predictions for all 9B single nucleotide variants, but can you trust them? The models are still far from accurate!
Check out our updated preprint (now a full paper, upgraded from our early pre-CSHL AI symposium draft):
biorxiv.org/content/10.64898…
Valli Subasri retweeted
1/ Most genes and variants matter somewhere but look silent in any one cell model. Can we build conditions that reveal phenotypes? With @kanishkadey, we engineer cell states to expose hidden disease biology and trace how genetic risk reaches disease genes. biorxiv.org/content/10.64898…
ALT Visual abstract: activating TFs singly and in pairs engineers new cell states, and the Mosaic model learns how TFs combine. Findings: hidden AMD risk variants, state-dependent convergence, and chromatin priming.
Valli Subasri retweeted
Training classifier-free guided flows? Global Transport keeps high guidance scales stable 💪 The trick ... compute mini-batch OT couplings without labels! Nothing else changes, so it’s simple and easy to use (and it works 🎯).
Check out the thread 🧵 for details.
Valli Subasri retweeted
1/ Purified at a spring, the high priestess Pythia entered a trance state, speaking prophecies that priests interpreted. Our Pythia predicts where RNA-binding proteins bind and we wanted to understand why.
Our preprint is out 👇
biorxiv.org/content/10.64898…
Valli Subasri retweeted
Delighted to see our real-world clinical study of a conversational diagnostic AI system published in @TheLancet, alongside coverage in @nytimes last week.
Looking back, AI in medicine might feel inevitable but it actually takes deliberate, uncertain leaps to get there. I have been fortunate to be part of two such moments.
The first was early 2022 (pre ChatGPT) when @alan_karthi and I got access to our first @Google LaMDA and PaLM models. That convinced us to go all in on LLMs in medicine, sparking the Google Brain moonshot leading to Med-PaLM.
The second moment leading directly to this Lancet study, was late summer 2023.
@taotu831 showed me the early AMIE models he was training with our self-play loop, targeting our upcoming OSCE study with patient actors, later published in @Nature 2025.
At that exact time, I was navigating a personal health condition. Testing myself, the consultation the AI performed was so natural and clinically sound that I was convinced it was ready for the real world.
So I reached out to Alan and together with the team we asked how we could test this in the real world quickly, safely and responsibly. We needed a prospective clinical trial with a safety endpoint and all roads led to @AdamRodmanMD at @bethisraellahey @harvardmed .
Adam was incredibly gracious with his time kicking off our first discussions in the fall of 2023.
But clinical trials take a long time; years of painstaking work. There are so many critical people and moments that make them happen, far beyond the glitzy PR materials and the main authors.
- For example, a discussion with @greg_corrado sparked the design of the oversight setup underpinning the safety layer in the study.
- Kavita Kulkarni (now at BioHub) spent months negotiating the initial contracts with BIDMC; it was uncharted territory for both organizations.
- @_khaledsaab (now rocking at OpenAI!) working tirelessly to improve the AI, together with @taotu831, @apalepu13, @RyutaroTanno, @ckbjimmy, @valentinlievin, @elahevedadi and many more folks.
- @Mysiak, Alessio Orlandi, Petar Sirkovic and team sharing the seamless serving infrastructure they built for Co-Scientist, which we could re-use for much of this study.
- @SSaraMahdavi making sure the study team was well organized :)
This is just to name a few. Prospective trials are a massive team sport and I am proud of what our teams achieved.
While the AI's clinical reasoning was strong, seeing the patient trust, safety, and satisfaction data, with ZERO safety stops from our supervising physicians, is what truly stands out.
At the same time, this is still a stepping stone. We are moving from AI as a tool to an active care team member - a guardian angel amplifying clinicians and staying with patients throughout their health journey.
More to come soon, and as always, a pleasure to work with so many great folks across @GoogleDeepMind , @GoogleResearch and @bethisraellahey .
Paper - thelancet.com/journals/lance…
NYT - nytimes.com/2026/10/08/techn…
Valli Subasri retweeted
This week we're releasing a preprint exploring state space models for long-sequence genomic prediction, alongside Cerberus, a set of open PyTorch models built on this work. biorxiv.org/content/10.64898…
Valli Subasri retweeted
#SpliceAI2 technical deep dive: transcript prediction and training. We perfectly reconstructed the top transcript for 82% of held-out genes. How? Splice graphs, dynamic programming, and long-read RNA-seq. 🧬✂️🧵
Valli Subasri retweeted
Incredibly, incredibly excited to introduce SpliceAI2, the next generation of splicing prediction! 🧬🚀🧵
Valli Subasri retweeted
Meet our lightweight (looped) tabular foundation model for bulk transcriptomics data!
Amazing work by @TheusResearch, Michael Bohl, and David Wissel.
We benchmark our GeneICL model on 80 classification, regression and survival prediction tasks and demonstrate its accuracy and efficiency!
Self-supervised transcriptomic FMs are pretrained on millions of gene-expression profiles, yet often fail to beat simple supervised baselines on clinical tasks.
Why?
Our new work suggests the pretraining objective may matter more than scale.
📄arxiv.org/abs/2610.08694
1/8
