Coding, app-building, AI

[kidthor]

Advertisement - Members don't see this ad

Hi all - I love my clinical work but I keep wondering whether I should learn more about coding, app-building, and AI going forward - as a supplement / side-gig option to actual anesthesia work. Honestly I think most of AI in healthcare is nonsense hype and fluff, but I'm wondering if it's worth learning more about - as I'm sure we'll be using some of it soon. It also seems like most in healthcare have at best a superficial understanding of these things (and sadly that includes all the IT/tech people). Most of the papers regarding AI in healthcare also seem like fluff proof of concept without implementation (e.g. they just plugged a dataset in to chatGPT to see whether it could theoretically predict sepsis on a different data set, and such).

Are any of you actually coding, building apps, or doing anything meaningful with AI? Anyone using Python or Claude Code?

I'd be curious how you'd recommend getting started learning in these areas and also curious whether you think these skills are actually applicable in anesthesia/healthcare?

 

[thinkorswim]

Do it if you’re genuinely interested, but don’t expect any meaningful income from it.

The entry-level software engineering job market these days is terrible. People who are already proficient with coding aren’t getting jobs or making money from it.

 

[kidthor]

Do it if you’re genuinely interested, but don’t expect any meaningful income from it.

The entry-level software engineering job market these days is terrible. People who are already proficient with coding aren’t getting jobs or making money from it.

I'm thinking about it more from a physician with added side-knowledge perspective - and out of general interest. IMO there aren't many who actually understand the intersection of healthcare+coding/tech. But yes transitioning to a full-time coder would be a fool's errand these days.

 

[T-burglar]

I have already made a large time and effort investment in learning software engineering from the ground up with the intent to maybe make something with an impact someday , even if it’s intra institutional.

To do it correctly is a significant investment. The tools available can get extraordinary work done for you but only if you understand a lot of fundamental programming and systems design already.

In lower stakes areas, like scamming people on the internet with a buggy garbage app, you can get away with less. If you want to build something for healthcare, you’re already targeting the most rigidly regulated area where you truly need an engineering mindset over a vibe coding mindset.

How to get started? Do a few rigorous online courses on the fundamentals of programming , web technologies , the browser as a shell for applications , etc.

 

[T-burglar]

Then move on to programming audit trails, BAA cybersecurity agreements for ePHI, the FHIR AND HL7 interoperability standards for healthcare data lol

Not trying to be an ass, just answering your question in an honest way. It’s *a lot* and this doesn’t even begin to cover it.

Try some courses on HTML, CSS, JavaScript, and the React framework and see if it still interests you after those. I’ve been through many many courses but the ones I took that made me feel like I could maybe do this after all were Brad Traversy’s introductions to HTML, CSS, and JS

And you’re right. Hardly anyone can speak the language of both tech (at the level of a programmer) and medicine. Not even the informaticists really have this combination of domain expertise. Learning this stuff is a slog sometimes but so was learning medicine. If you’re comfortable with your clinical skills and looking for a new challenge this is a decent choice imo. Either that or pick up gardening

 

[kidthor]

@T-burglar - that is all quite helpful, thank you.

I've failed at gardening, so many this is a reasonable side project.

What do you think the deal is with all the people doing "AI research" that consists of superficial Claude/Chat-GPT queries and then giving talks on it at a conference. Do those people have any there-there or is it all fluff? Also what's the deal with all the hospital IT/tech people who don't really seem to understand anything either? Are they just collecting a paycheck, going to meetings, and pushing paper while Epic/Cerner maintains the shabby (and exorbitantly expensive) status quo?

 

[T-burglar]

There’s “AI research” which is people being paid to talk to the chatbot - almost pointless. Then there is AI research science which is all data structures and algorithms and machine learning math.

This is a novel and impactful new technology but much like most things in the corporate world everyone is putting in zero effort to really understand it. Yes the vast majority of it is bull**** and fluff

 

[T-burglar]

I basically do all my database schema design and implementation through Claude code now (postgreSQL) and just spot check it with a test harness. Ever since opus class models have been out I have not seen it miss the mark - and the models will only get better. I haven’t even used Fable yet even though I have access because opus high seems to be fine for my purposes. Sometimes I even still use sonnet

 

[GassYous]

Theres an em doc groove who is using ai for investment strategies and seems to be doing really well. I think he was a programmer in his previous life.

 

[T-burglar]

I have a lot of thoughts on this subject so I apologize for spamming the thread. If the idea of learning programming principles and practice is daunting to you, perhaps educating yourself conceptually on how this technology works would be more reasonable . It’s a more manageable task I think than learning to program as a mid career physician (I am nuts and an outlier, I know).

With enough understanding, you could position yourself maybe to be one of the people in your shop doing paid non-clinical work involving their implementation. Or my own personal prediction about what new positions will become available in the near future - developing, validating , and maintaining custom intra institutional SLMs to contain costs , wall off institutional trade secrets from the web, and get most of the functionality of the frontier models.

 

[sloh]

AI is quite useful on the compliance side for providers if you are interested in a role like that. It really requires a physician as a human in the loop to integrate the clinical side with regulations, coding, and billing. And increasingly, it can't be done to the same quality or speed without knowing how to use AI.

 

[Twiggidy]

I’ve tinkered with Python a little and have recently thought about really diving in and learning more maybe even pursuing a certificate. Of late I’ve been researching if there’s just little ways knowing some coding can make some things just easier for me personally ie any sort of DIY type stuff for personal uses. I’ve had some medicine related ideas but have no idea how to implement them. I figured to just start with a couple workbooks and see what happens because it’s not a giant investment

 

[pgg]

I’ve tinkered with Python a little and have recently thought about really diving in and learning more maybe even pursuing a certificate. Of late I’ve been researching if there’s just little ways knowing some coding can make some things just easier for me personally ie any sort of DIY type stuff for personal uses. I’ve had some medicine related ideas but have no idea how to implement them. I figured to just start with a couple workbooks and see what happens because it’s not a giant investment

Once you get the very basic stuff down and can write and compile and run a project that's more complicated than Hello World, if you decide to formalize your learning a little, consider taking one of the free online university level courses in data structures and maybe algorithms. Something like these -

Data Structures

Offered by University of California San Diego. A good algorithm usually comes together with a set of good data structures that allow the ... Enroll for free.

www.coursera.org

Algorithms, Part I

Offered by Princeton University. This course covers the essential information that every serious programmer needs to know about algorithms ... Enroll for free.

www.coursera.org

Data Structures and Algorithms

Offered by University of California San Diego. Master Algorithmic Programming Techniques. Advance your Software Engineering or Data Science ... Enroll for free.

www.coursera.org

When I was an undergrad I spent a couple years working in a lab, kludging together data collection programs in C that was mostly self-taught. Programs compiled with lots of warnings 🙂 but mostly ran fine and didn't crash often. About halfway through that, I started taking comp sci courses toward a minor. Entry level ones were just very basic languages and basic algorithms.

I ended up taking 5 or 6 upper division courses to finish the minor, but the ONE course that helped me move from a "can get it to work, eventually" programmer to one who was "competent" was the introductory data structures course. It literally changed the way I thought about coding, from planning to actually writing them. There are so, so many things novices do when writing programs that make simple, logical, intuitive sense but are just plain wrong, buggy, slow, or dangerous.

No you'll probably never code a binary tree or need a hash table (nor should you, except for fun - if you need one, you should use an existing library). But it's worth understanding what they are and why you'd want one and how to use libraries that do those things for you. And if you're a bit of a nerd and find this stuff interesting, it's fun.