Julia's biggest success[es] so far?

Community
1

I’m thinking of a) [enabling] commercial success and/or b) research success and/or c) good behind-the scenes package, likely at 1.0+. [And maybe Julia website in production?]

For a) you might not know of a specific company, and nothing might be obvious from looking at the public package repository.

However, I believe Pumas.AI (I believe part of COVID vaccine success) might be the largest success so far, a company built on SciML. Any other ideas?

DifferentialEquations.jl enables SciML, and is now under that umbrella. Note, I do not think general machine learning (e.g. Flux.jl) is yet a big success for Julia in the commercial sense, and even for use in research, could be wrong, at least not the mainstream SOTA research I follow. SciML is however not typical ML, and is a huge success.

SciML is a huge ecosystem, some of it rather new, e.g. ReservoirComputing.jl not yet at 1.0.

The first state-of-the-art Julia package was I believe JuMP.jl, but only recently after 10 years hit 1.0. I count it under b) research success and c), hopefully helped some companies too, likely wouldn’t/couldn’t know easily.

Genie.jl for building websites is now at version 4.18.0 whatever you want to read into that. [EDIT: It’s claimed ready for production, see from the horses-mouth below, so striking out repeatign incorrect claim to the contrary.]. It is used in production for intranet sites (for “billion dollar companies”), and also publicly open website.

Julialang.org however uses Franklin.jl for static website. It’s the most high-profile website run (in some sense) by Julia.

DataFrames.jl is now at version 1.3.3 a good candidate for c).

Is there a good way to filter the General registry for packages already at 1.0+? Anyone has code or can come up with, or maybe possible at juliahub.com?

Other Julia success companies I can think of:

Invenia is also a company built on mainly I think Julia with an impressive github: Invenia Technical Computing · GitHub

RelationalAI (likely not yet a commercial success).

Blackrock (not built on Julia originally, but adopted Julia, still using?), largest such company. Julia is also used in other finance companies, likely too new to for companies to build on it only.

4 Likes
2

Can you please share the Reddit thread where I’d claim that Genie is not production ready? That’s certainly not true. I have personally developed quite a few apps for various consultancy clients and institutional customers, some of them being in production for years now (mostly backend/intranet apps though). In addition, I have collaborated with users and contributors that deploy Genie apps at work, some working at billion dollar companies. Not really at liberty to share details though, as it’s usually the case.

We’re now starting to put more effort to feature user apps on GenieFramework.com - but for an example of a great recent app that’s been running in production for over 3 months, with public access and public repo, people can check https://pkgs.genieframework.com for a good example of:

  • reactive data dashboard
  • with REST API
  • with SwaggerUI API
  • currently crunching 1,700,545 rows of package downloads data to compute stats in real time
  • on a puny 1GB RAM Heroku server (with some 1/2 of that RAM available)
  • with avg server response time of 500 ms
  • featuring lots of best practices including container/Docker deployments with prebuilt Julia sys img, caching, CDN support, MVC architecture, DRY, etc.

Version 4.18 means 4 breaking versions so far :slight_smile: But more importantly it’s about over 3-4 years of continuous improvements, over 1500 commits and some 60 awesome contributors. In addition, Genie 5 will be released in a couple of weeks, as our first performance focused release, which should address the concerns of some of the people in the Julia community around Genie’s speed.

25 Likes
3

Well he wrote Genie “author” (I assumed you) claimed it and I responded, and he did yesterday respond repeating it for “Genie fella”, so it might be better if the denial came from you:

https://www.reddit.com/r/Julia/comments/u8shoe/state_of_web_assembly_in_julia/i5n5l5w/?context=3

4

Cool, thanks. Who knows what he meant/understood. But yes, I’ll follow up, worth clarifying.

5

Here’s a short testimony by @Satvik about a commercial success that I bookmarked last year. It’s (probably deliberately) vague about the details but I found it impressive and convincing nonetheless.

4 Likes
6

There’s probably about a billion dollars worth of predominantly Julia-based companies:

Putting on my Julia Computing hat, many more success stories using the language appear on our case studies page — even for folks that aren’t directly working with our company.

The foremost research project in my mind is the Clima project.

20 Likes
7

Thanks for the mention! We’ve continued to move a lot of our Python code to Julia since then (and to write a lot of new Julia code), and have generally found that our average Julia code is 2-5x faster and more memory-efficient than our most optimized Python/numpy code.

One of the biggest gains was that Julia’s macro system allowed us to switch from calculating our features directly to building and evaluating a computational graph, which allowed us to skip a lot of redundant calculations. The fact that Julia’s for loops are fast also made it easy for us to do incremental updates, whereas in Python we’d essentially had to settle on batch refreshes every time we wanted new data.

I’m happy to go into more detail on any of these points, I just don’t know what people are interested in.

18 Likes