Oh, yeah, game mechanics or game physics engineering is amazing, but understand it's about 1% (if that) of what makes a successful game. I was watching my nephew play Fortnite over thanksgiving and thinking to myself "gee, the graphics are shit, essentially low-poly with some nice shaders!". And CoD Black Ops puts the physics to shame (and thinking that old timers used to tell me the physics of that were bad compared to e.g. PubG/Battlefield). But Fortnite is super colorful, and for a kid just coming off Roblox, it's accessible enough to play. So the HUDs and UIs and leveling and map dev and a million other admin tasks are where the majority of your game dev experience will be. Even doing some (gasp!) web dev, for things like in game stores and so forth. The gameplay mechanics and physics math of particle engines or calculating missile intercepts or inverse kinematics (jacobian matrices anyone?) for skeletal animations are mostly solved problems, but they can still be interesting to reimplement.

To me it sounds like someone's still making a living from R&D even though the big money is at casual games and selling diamonds or hats in in-game stores. The question is just how to get that job.

Back when there was no commoditization of game engines, every team starting a game had to do their own engine, tools etc.

Now that you have unreal and unity etc, you can actually concentrate on game design and gameplay instead of worrying about the engine and tooling.

I agree though writing a game engine and physics engine is very interesting. I used to work for the R&D team of a casino games manufacturer. We created our own in house game engine. It was fun, but adding any new feature would take a very long time. In the end we shifted to unity.

Wouldn't you just use an off-the-shelf rendering and physics engine instead of working on your own?

indeed, Fortnite uses Unreal engine

More accurate to say that fortnite is developed alongside unreal engine, they make unreal engine in house so they make their own engine. For example unreal engine 5.0 was very buggy, since they made it to work with fortnite first so only fixed the features that game used.

there would be some crossover but I'd imagine they're two separate entities

I've made my own simple physics engines and used linear algebra to do things like projecting from 3d to 2d space. None of it was any good but it was a lot of fun. This was back when I was most comfortable in Java and the performance was horrible. I don't know what methods of collision detection real engines use but Runge–Kutta is not a good option for real time games.

I don't know how to do it and make money but I would recommend playing with the math. It's like magic, you setup the formulas and things just start moving on the screen. It was so much fun.

Actually there’s still quite a bit of math involved implementing gameplay mechanics and it can be quite a bit more off the cuff and experimental which people might find fun. The equivalent of writing shaders rather than a rendering pipeline.

Exploration geophysics, remote sensing, satellite navigation, etc.

Fugro, Siemens, et al - Earth visualisation, medical imagery, etc.

The emphasis in these fields is getting visualisations and computations correct rather than convincing.

ie. The math part is interesting.

In remote sensing you might find yourself doing a continous pipelined SVD reduction of 500+ dimensions from a spectrum sensor to some three axis RGB like space to create a viewable image from a non visible part of the EM spectrum.

There's not so much of taht in game dev.

Any chance of doing that while not working in a giant crappy corporation, like Siemens?

Sure, I spent most of my career working for small nimble outfits that ultimately sold product to large crappy companies while moving on to other challenges.

The pipelined SVD reduction mentioned was written while working with a small company with 13 airframes (fixed wing and helicopters) that did full spectrum rad+mag+grav+lidar surveys of small country sized areas (Fiji, Mali, other similar sized regions) and eventually sold to Fugro.

Sometime before that there was something that was aquired and became [1]

[1] https://hexagon.com/company/divisions/safety-infrastructure-...

but that was very pre google earth | maps.

Today there are oppotunities such as [2]

[2] https://www.abc.net.au/news/2023-01-08/space-satellite-image...

If you think that the math part of game engines is interesting but aren't that interested in promoting an indie project, then you should look for jobs in robotics or computer vision applications. There isn't a large supply of jobs for replicating what can be handled by Unity or Unreal Engine.

I recommend medical imaging as a subset of those. But perhaps unlike robotics and computer vision, there’s less possibility to learn as you go.

What is it like to work in computer vision these days? I did some courses back in the day and it was interesting, but to me it seems that it's now devolved to looking up convolutional DNN architectures and trying which one works best for the problem at hand.