Cyril Zakka, MD (@cyrilzakka)

Pinned

C

Cyril Zakka, MD@cyrilzakka · Nov 29, 2023

There’s been a lot of buzz surrounding multimodal LLMs in medicine. At the @HiesingerLab we’ve been building Almanac Chat for over 6 months alongside our collaborators and we’re excited to showcase the first preview! • Almanac Chat is the first truly multimodal medical LLM…

16

56

345

199

84.0K

Cyril Zakka, MD Retweeted

D

David Ouyang, MD@David_Ouyang · Jul 24

We are excited to work with @StanfordDeptMed, @SmidtHeart, @mghliver, and @KPDOR with the support of @American_Heart to refine and prospectively evaluate AI opportunistic screening approaches for Cardiovascular-Kidney-Metabolic Health! Chronic Liver and Kidney Disease is common…

4

19

76

11

7.0K

C

Cyril Zakka, MD@cyrilzakka · Jul 23

A lot to unpack here but having open-source and open-weight AI as core tenets of the American AI Action Plan is going to have compounding benefits for the quality and affordability of healthcare in the US. We’re still in the early days but a shift is coming. 🇺🇸🤗

SSriram Krishnan@sriramk · Jul 23

🇺🇸 Today is a day we have been working towards for six months. We are announcing America’s AI action plan putting us on the road to continued AI dominance. The three core themes: - Accelerate AI innovation - Build American AI infrastructure - Lead in international AI…

2

5

32

2

3.0K

Cyril Zakka, MD Retweeted

S

Sriram Krishnan@sriramk · Jul 23

Tomorrow should be a huge day for American AI. 🇺🇸

97

129

2.0K

168

542.0K

C

Cyril Zakka, MD@cyrilzakka · Jul 17

There’s recently been an influx of case reports describing people exhibiting signs of psychosis having their episodes and beliefs amplified by an LLM (e.g. ChatGPT, CharacterAI, etc) While I’m not a psychiatrist by training, I think it mirrors an interesting syndrome known as…

7

85

32

10.0K

Cyril Zakka, MD Retweeted

G

Georgia Channing@cgeorgiaw · Jul 16

You know CASP? The competition that AlphaFold won that changed the game for AI x bio? 🧬 Just dropped all the data from their last challenge on @huggingface! check it out ⤵️ huggingface.co/datasets/cgeor…

1

9

34

10

2.0K

C

Cyril Zakka, MD@cyrilzakka · Jul 14

and it's out!

�🍉 Abubakar Abid@abidlabs · Jul 11

🧑‍🍳 We're cooking up something. You'll soon be able to go from any API with an OpenAPI spec to a Gradio UI. Coupled with @Gradio's MCP integration, this makes it especially easy to go from API -> MCP.

1

3

23

5

4.0K

C

Cyril Zakka, MD@cyrilzakka · Jul 9

So many great releases today, with a new medical model to boot! 27B multimodal MedGemma featuring my all-time favorite encoder: huggingface.co/papers/2507.05…

OOmar Sanseviero@osanseviero · Jul 9

🏥Introducing MedGemma, part 2, including: 🔥A 27B multimodal MedGemma 👀MedSigLIP, a lightweight image/text encoder for medical image retrieval/classification 📜A technical report with details Blog: research.google/blog/medgemma-… Paper: arxiv.org/abs/2507.05201

0

2

21

6

4.0K

Cyril Zakka, MD Retweeted

A

Abhinav Adduri@abhinadduri · Jul 9

We updated the State Embedding 600M checkpoint on the @ArcInstitute Hugging Face This model was trained with 4x FLOPs compared to the preprint model. It achieves significantly lower val/loss and does better on internal evals - would recommend using this over the 4 epoch one for…

1

9

52

9

15.0K

Cyril Zakka, MD Retweeted

M

Michael Moor@Michael_D_Moor · Jul 9

Want to do a postdoc in my research group? bsse.ethz.ch/mail I can support one ETH Postdoc fellowship application this fall. For details, see: grantsoffice.ethz.ch/funding-opport… If interested, please carefully check your eligibility (esp PhD defense date) and then send us the…

1

10

29

3

5.0K

C

Cyril Zakka, MD@cyrilzakka · Jul 8

We just released the best 3B model, 100% open-source, open dataset, architecture details, exact data mixtures and full training recipe including pre-training, mid-training, post-training, and synthetic data generation for everyone to train their own. Let's go open-source AI!

LLoubna Ben Allal@LoubnaBenAllal1 · Jul 8

Introducing SmolLM3: a strong, smol reasoner! > SoTA 3B model > dual mode reasoning (think/no_think) > long context, up to 128k > multilingual: en, fr, es, de, it, pt > fully open source (data, code, recipes) huggingface.co/blog/smollm3

42

244

2.0K

795

188.0K

Cyril Zakka, MD Retweeted

L

Loubna Ben Allal@LoubnaBenAllal1 · Jul 8

Introducing SmolLM3: a strong, smol reasoner! > SoTA 3B model > dual mode reasoning (think/no_think) > long context, up to 128k > multilingual: en, fr, es, de, it, pt > fully open source (data, code, recipes) huggingface.co/blog/smollm3

69

203

1.0K

524

266.0K

Cyril Zakka, MD Retweeted

V

Victor M@victormustar · Jul 1

Sad news: HuggingChat is gone 😢 It was a wild adventure... 🔹20+ Open Source models 🔹1M+ users 🔹100k+ assistants created But with where AI is going lately, it was starting to feel out of date. It’s time to build something new. Stay tuned…

25

14

201

24

15.0K

Cyril Zakka, MD Retweeted

�

🍉 Abubakar Abid@abidlabs · Jun 30

Why would you create an MCP server?? When you could create a UI, API, and MCP server all with the same Python code?

3

6

57

28

17.0K

Cyril Zakka, MD Retweeted

V

Vaibhav (VB) Srivastav@reach_vb · Jun 30

no matter when it drops it'll be at hf.co/openai

0

2

18

2

2.0K

Cyril Zakka, MD Retweeted

M

Michael Moor@Michael_D_Moor · Jun 30

🚨New preprint! 🚨In-context learning (ICL) is the intriguing ability of LLMs to learn to solve tasks purely from context w/o parameter updates. For multimodal LLMs (MLLMs), ICL is poorly understood, especially in the medical domain where doctors would often face few relevant…

2

20

86

49

7.0K

C

Cyril Zakka, MD@cyrilzakka · Jun 26

I f*** love this. And to put even more momentum behind this awesome movement by our @arcinstitute friends: 🥁@tahoe_ai will give $25K to the best model that is also open-sourced with weights on @huggingface for everyone's use! + maybe a mention on X by @ClementDelangue? ;)

AArc Institute@arcinstitute · Jun 26

Register today for the Virtual Cell Challenge and use AI to solve one of biology’s most complex problems. Announced in @CellCellPress, the competition is hosted by Arc Institute and sponsored by @nvidia, @10xGenomics, and @UltimaGenomics.

3

12

72

15

26.0K

C

Cyril Zakka, MD@cyrilzakka · Jun 23

Great new model from the @arcinstitute. Model links here: huggingface.co/arcinstitute/S…, huggingface.co/arcinstitute/S…, huggingface.co/arcinstitute/S…

AArc Institute@arcinstitute · Jun 23

Introducing Arc Institute’s first virtual cell model: STATE

1

14

3

7.0K

C

Cyril Zakka, MD@cyrilzakka · Jun 20

Excellent commentary, @cyrilzakka. Fascinating implications for teaching residents. @TradIMYale

CCyril Zakka, MD@cyrilzakka · Jun 19

There's a lot to be said on this newest preprint showing that reliance on LLMs in writing tasks leads to a 'consistent underperformance' in neural and behavioral levels with potential long-term consequences. Interestingly, I think this ties to similar but (early)…

0

1

3

1

2.0K

C

Cyril Zakka, MD@cyrilzakka · Jun 19

Very good point. But the cognitive impact of technology is a recurrent question over at least the last 75 years (more recently for Search—see article below).

CCyril Zakka, MD@cyrilzakka · Jun 19

There's a lot to be said on this newest preprint showing that reliance on LLMs in writing tasks leads to a 'consistent underperformance' in neural and behavioral levels with potential long-term consequences. Interestingly, I think this ties to similar but (early)…

5

8

29

12

6.0K

C

Cyril Zakka, MD@cyrilzakka · Jun 19

Unsurprisingly @cyrilzakka says it better than I could. These are the exact concerns I have about the way we are integrating LLMs into clinical care. It *might* not matter, but we are all participating in an educational Pascal's wager right now. A 🧵🔽

CCyril Zakka, MD@cyrilzakka · Jun 19

There's a lot to be said on this newest preprint showing that reliance on LLMs in writing tasks leads to a 'consistent underperformance' in neural and behavioral levels with potential long-term consequences. Interestingly, I think this ties to similar but (early)…

1

18

47

24

11.0K