Roxana Daneshjou MD/PhD

Assistant professor of Biomedical Data Science @StanfordDBDS and Dermatology @StanfordMed | AI/ML & precision health | @Rice_BioE alum | @pdsoros Fellow 2014

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Excited to see our @NatureMedicine paper out today led by @ChanwooKim_ with my co-senior author @suinleelab and an amazing team! We used a dermatology foundation model to enable explainable and transparent AI - from auditing datasets to models. https://t.co/KfqjMV9j5A

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

@EricTopol @DShaywitz @zakkohane @NEJM_AI I suggest you look into since many of the major AI in healthcare papers get published there. Hard to keep up in the field without that journal on your reading list.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

@kvnwang @ritwikpavan @christyclane @emilykrausmd They’re going to put out a technical video very soon. It’s multiple sensors with IP.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

@kvnwang @ritwikpavan @christyclane @emilykrausmd They use multiple physiological sensors + ML + calibration per user. I’ve seen the prelim studies comparing to urine measurement and it looks good. Cofounder previously successful in noninvasive monitoring space as a CTO.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Published in Annals of Internal Medicine: https://t.co/bdQAaeJ4ix

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

What they fail to mention here is that ChatGPT often gives wrong health advice, including a case where a man ended up with bromide toxicity thanks to ChatGPT advice.

[OpenAI @OpenAI]

Every day, millions of people ask ChatGPT about their health – from breaking down medical information, preparing questions for their doctor's appointments, to helping people manage their overall wellbeing.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Let me point this out in case you missed it. NEJM AI, one of the top journals in the field, is showing how AI can be safely leveraged in peer review. In this case, there was a human editor in the loop and statistical review was also done by a human.

[NEJM AI @NEJM_AI]

Editorial by A.K. Manrai (@arjunmanrai) et al.: Accelerating Science with Human+AI Review https://t.co/Cd7zuiD7Md

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Be careful with using AI to craft your personal statements. AI-based statements begin to sound REALLY similar (the tone, wording, etc). This will cause your readers to zone out. The fresh voice of an AI-less personal statement, on the other hand, will snap them out of it.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Excited to share my research highlights for 2025! Check out this personalised summary of my Springer Nature publications and the impact of my research in 2025. https://t.co/TphB7scc4B #SNAuthors #YiR2025

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

@joejanizek It’s just the absolute worst 😂

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

-Not outlining your specific research contributions in SOP
-Getting rec letters from people who don’t know you well. Do not pick famous people over people who know you well.
-Not asking, “can you write me a strongly positive letter?” from your writers. Tepid letters are awful.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Common mistakes to avoid in your PhD applications (from someone who evaluates them):
-Mostly mentioning faculty from other departments (instead of the one you’re applying to)
-Having a statement of purpose that doesn’t lay out your research direction (we know it might change)

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

@morgancheatham @BrownUniversity Just so so awful.

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Seven year old just said to her dad, “I don’t want your idea to be part of my project. I want to be independent.” 😂

[npj Digital Medicine @npjDigitalMed]

Generative AI models are giving fewer medical disclaimers over time. 📉

In 2022, ~26% of AI health answers had a disclaimer.
By 2025? <1%.

As models get smarter, they’re getting less safe. Patients may take outputs as medical advice.

https://t.co/2OYQvKdezT

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

It’s clear that despite what they say, AI companies are actively going after consumer health (with few guardrails). See our work below:

[npj Digital Medicine @npjDigitalMed]

Generative AI models are giving fewer medical disclaimers over time. 📉

In 2022, ~26% of AI health answers had a disclaimer.
By 2025? <1%.

As models get smarter, they’re getting less safe. Patients may take outputs as medical advice.

https://t.co/2OYQvKdezT

[Tom Hartvigsen @tom_hartvigsen]

📢Please retweet: We are hiring a **Postdoc** at UVA to work on Continually Monitoring and Updating Multi-modal Medical AI Models!

Great opportunity to design impactful methods alongside great collaborators @_ahmedmalaa and @RoxanaDaneshjou

More info: https://t.co/a3Qs5X0pgy

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Paper: https://t.co/H5OOIE9lbA

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

We know patients are asking medical questions from public facing large language models. And previously these models would often give a disclaimer that the patient should speak to their doctor. But these disclaimers have decreased across ALL models!!

[Roxana Daneshjou MD/PhD @RoxanaDaneshjou]

Game over. The 21 month old knows what tahdig is and demands it by name when he sees me having a piece. Between him and the six year old, my tahdig eating days are over.

[Ahmed Alaa @_ahmedmalaa]

Excited to launch our project on clinician-in-the-loop post-training methods for medical multimodal AI with @RoxanaDaneshjou and @tom_hartvigsen. Grateful to @NIH for supporting this work!