As someone whos has written a very large and complex toolbox in Matlab [1], and now rewriting in Python I would not recommend locking yourself in to Matlab/Mathworks. However, if you don't use any GUI functionality you might get away with it, and you might be able to use Octave which can run must Matlab code unmodified (also see Octave online for deployment).

[1]: https://www.featool.com

Thank you very much for your advice. Is there any specific reason to rewrite to Python?

Sure. As you guys seem to plan to do, I also started with some Matlab based work from my PhD years and felt that would be the quickest and easiest way forward. I also made the bet that the Octave app GUI functionality (uixx functions) would become good enough to use for commercial apps, which as great as base Octave is, it in my opinion hasn't. So as I make a fully client side app I am now dependent on users also purchasing or already having Matlab too.

As for Python, it is also a bet and compromise with the following specific criteria:

- Fast and easy to develop in (as I as a bootstrapped solo-dev have to do everything by myself)

- Reasonably simple to translate to from Matlab code

- Good and easy to use client side GUI functionality

- Easy for other users to use (as I want as many users a possible to use the functionality and potentially be able to write extensions themselves)

- 3D Visualization libraries (VTK, ParaView)

with this in mind my choices were:

1) Python

2) Julia

3) (Object) Pascal

4) Web stack (Javascript + webassembly)

I almost initially chose (Lazarus free) Pascal due to its IDE, fast compilation, and easy deployment but eventually I found the fixed pixel based GUI placement not ideal (although there are probably workarounds for that). Although using a web stack would be ideal from a user and sales perspective (with no installation), my impression of web programming is "messy" and I'm afraid that for CAE simulation apps this approach would freeze the computers completely (Electron!). Although I initially thought Julia would be ideal as the syntax was very close to Matlab, it seems to become more and more complex as it matures and my impression is that there still isn't a good way to deliver (pre-)compiled client side apps which is my requirement. So all in all I felt that Python probably the best compromise. It is still a bet, but with the "Matlab engine" you can call Matlab in the background so port bit by bit. In your case though with a (web) server-client app you might fare better with Matlab/Octave.

Tech-stack aside, I've found the hardest part is the non-techy bits, especially sales, marketing etc. which I still haven't figured out.

Best of luck!

MATLAB has "export to C" capability:

https://www.mathworks.com/help/dsp/ug/generate-c-code-from-m...

Build your website however you wish, then call the C code (e.g. C# P/Invoke interop if you use .NET, Python CFFI, FastCGI, etc.)

> Run Matlab Production Server on AWS as a microservice. But the licensing cost of Matlab Production Server is "contact us". Scary.

Starting a business is scary. Contact them, figure out the cost, and figure out if it's worth it to rewrite it in a freely licensed language.

Contact them and see if it really is as bad as you think :)

You can also go self-hosted while you are in customer acquisition mode?

If the cost of a software license is scary, then I can assure you labor costs are much worse. You can rewrite now or rewrite later, but hiring devs is a ways away :)