James Weimer

Congratulations to Profs. Kevin Leach and James Weimer for their exceptional research in software security and medical devices, which has secured them the prestigious $2 million ARPA-H contract! engineering.vanderbilt.edu/2024/05/21/van…

James Weimer, assistant professor of computer science and noted medical-device entrepreneur, has received a @NSF CAREER Award to improve the safety and efficacy of systems that use advanced machine learning and artificial intelligence techniques. engineering.vanderbilt.edu/news/2024/weim…

There are MANY unmet foundational scientific needs that can only be identified and effectively solved by getting outside the academic bubble and understanding real-world needs/constraints.

My NSF CAREER was awarded. nsf.gov/awardsearch/sh… A big thanks to the NSF CPS program and all my collaborators (especially the students) and my colleagues in @VUEngineering , @Vanderbilt_CS , and @ViseVanderbilt.

At the risk of being provocative, the support from @VanderbiltU since joining in '22 to pursue "radical collaboration" and interdisciplinary research has been beyond comparison. I hope more early career PhDs find home intuitions (like @VanderbiltU) where tech transfer is valued.

@KarlZelik I also have an app for any watch running WearOS that does something similar. Both watches can be configured to work for about 24 hours with constant sensing (10-30Hz). I used them for both my startups -- with clinicians seeing them up. Pretty bullet proof.

@KarlZelik Samsung Galaxy Watch Active with an app I wrote. I have 30 watches on campus if you need some for testing

🙋🏽‍♂️ Question for my research friends and wearable sensor gurus:  ⌚️ Are there any smartwatches (or wrist-worn activity trackers) that allow you to record and export continuous IMU data?

@jefielding I teach an Internet of Medical Things course in CS at Vanderbilt: covers engineering, clinical, and business aspects of IoMT (since you can't have impact in IoMT without all three working together) -- Do you have any modules related to Medical startups I could adapt (and cite)?

9% of my Startup Studio students thought they wanted to bootstrap their business at the start of class and 90 minutes later after my 📣 lecture on How VC Math Works, 33% prefer to bootstrap. First day of class, my job is done 😇

Want to be a CS researcher and also maximize the likelihood your work has real-world impact in medicine? Apply to join my phd research group at @Vanderbilt_CS -- We measure impact not only by "likes" (i.e., citations) on Google scholar, but more importantly by lives saved.

@MASI_Lab And Nancy is #1! Although your tag looks more fun. :-)

Our masi delegation had a great time at #ISMRM23. We shared excited results and came away with many ideas for new science.

@KarlZelik On the IoMT side, a corollary is consumer health information (CHI): unregulated personal feedback that provides general health information (but not diagnosis). FWIW, the CHI space has the same issues -- and no amount of screaming about bad behavior has changed things (I tried).

@james_e_weimer Clinical exoskeletons: Yes (they'd be FDA regulated) Clinical sensors: Sometimes Occupational, sports, recreational exoskeletons & sensors: No (generally they're not treating or curing; often they are for monitoring or prevention only) Half or more exos are not medical devices

🚀 5 years ago, I stumbled—somewhat unexpectedly—into the world of tech translation, commercialization & startups. It's been a wild & enriching adventure! 👨🏾‍🔬 As a scientist-turned-entrepreneur, this journey opened my eyes to the business world. But...

VasowatchAI is released (jamesweimer.net/VasowatchAI)!!! Model: predicts postpartum hemorrhage before delivery from maternal heart beats; 4x better than SOTA; and only 73 parameters because we spent 1.5 years studying physiology before model training. Next step: regulatory pre-sub!

Are you working in explainable ML/AI for hypothesis generation in life sciences? Ideally, you’re working within biotech, pharma, a startup, or a lab collaborating around life science discovery tasks. I have an ask for a NeurIPS talk but I’d love to invite others instead. Thanks!

@suchisaria I've noticed an uptick as well (but not 5-7 times a week!) I have capacity to take on more "discussion asks" in health AI. I have no agenda -- but believe a holistic approach is needed (starting from data generating devices, through training and beyond). Forward asks as desired.

There is an AI discussion ask — whether it’s from the gov, VCs, or National collaboratives — 5-7 times a week! Reminds me of the last AI wave… for a while, you had to make a decision about whether to do the work or inform the conversation. I like to, and I imagine many researchers, are prioritizing research/doing-the-work and so we are left with conversations often led by those less informed. There are so many amazing researchers & practitioners that I’d love if they were more involved! I’ve been forwarding asks I get … but need more leaders who care about the field/doing-the-right-thing over personal agenda (directing money to themselves / their institution / or becoming individually well-known). Do I sound like a cynic?

@NikolaiMatni I think I know this answer, but I'll ask anyway. How would one use this result in practice? What's the limitations of the beta-mixing assumption?

and thus exhibits graceful degradation as we introduce misspecification. Past a burn-in, our result is sharp in the moderate deviations regime, and in particular does not inflate the leading order term by mixing time factors.

🔥 Hot off the press 🔥 Who doesn't love least-squares, right?! Everything is 🎂 in the iid and linear setting, but what if you have dependent data, and even worse, you don't have realizability 😬 arxiv.org/abs/2305.11165

@suchisaria @bayesianhealth Sold. Just let me know if/when you need some IoT wizardry!

Headed to our 3rd full team retreat at @bayesianhealth 🌵🏜️💃 There will be rap battles, iron chef, dance competition, pool games, an award night and loads of team bonding over 4 days. I remember 3 years ago struggling to understand how to build a company with a great culture. I feel so proud of where we’ve come. This team feels so much bigger than the sum of its parts — they’re individually brilliant but highly collaborative. We genuinely care to solve the problems we’re tackling and have a track record of getting stuff done. I hate BS and so we avoid it. And we’re working towards a mission that’s so rewarding, work feels like play! Come join us if you’re looking for opportunities. We are especially looking for super ML & data engineers and data products folks who want to use their superpowers to improve the way health care is delivered. Email careers AT @bayesianhealth.com

@suchisaria Let's make this happen! As you suggest, building a community is definitely the first step. Take it offline to brainstorm?

@james_e_weimer I love the idea of partnering. We should pull together a solid group.

Served as a presenter/discussant at a brilliant White House discussion earlier today with heads from the FDA, ONC, HHS, VA and 30-40 representatives from different agencies. Q&A was focused on (a) understanding current state (b) will it raise cost (c) consent (d) regulation and infrastructure for reg (e) patient empowerment. Very level headed discussion. Interest in AI safety and reg is at an all time high. However, we need more individuals with deep knowledge of AI, experience translating solutions to the real world and regulation.

@nicolapaoletti @miladmkm Congrats! Very exciting! We need more MCPS research!!!