do you have a blog?
I'd be happy to go on about the project in as much detail as you please! Email me? Tried the address hinted at in your profile, but it bounced.
VR is a new medium. We're in that stage now where people are sort of putting stuff from the old medium in without much consideration of what new concepts or ideas might be possible. Every UI in every game and application is terrible. Everything has to be reinvented, and can be reimagined in incredibly creative and strange ways.
Fairly often now I'll have an experience in real life and think "Someday, I'm going to put this in VR game."
Anyway - I was tasked with removing Unity and refactoring the service locator usage, since Unity was going to be unsupported soon. Doing this gave me a real sense of agency in our otherwise boring enterprise app; it was one of the rare chances where I could make a significant, wide-scale improvement that would make future development easier for everyone.
(if it were me, i'd be hoping it was one giant monolithic application in one single repo, to make it easier to track & perform this kind of sweeping surgery)
Really enjoy the work itself, there’s tons of interesting algorithms/optimization involved
I feel I actually use my CS knowledge from university more doing this work
Its also really satisfying to visually see the result of your code in the game
On the other hand I do find it a lot harder of a job since you not only have to understand code but also the math behind the code
how was the transition?
Transition was definitely tough in the beginning due to steep learning curve
You think you know a decent amount from doing tutorials/side projects but a real production engine is a lot more complicated because its been optimized over the years and as a noob its not obvious why things are designed the way they are.
I felt like web dev you can become pretty competent in like 2-3 years but with graphics it feels like the ceiling is much higher and starts to verge on research as you get more into the topic
A lot of newer rendering techniques you can't easily google and you have to end up reading papers if you want to understand/implement them
Unfortunately, the computers used for the desk jobs are horrendously broken. Ie, takes ~30 mins to go from log-on to getting email up; everything has a long response time. Opening a PDF to sign or fill out takes 1-3 minutes or so each.
Yeah this is a ton of fun!
It's not every day you get the chance to just re-write something [it rarely is the best idea] but when you do, I relish it.
Still entangled in combining the different data sources before start building/training models. So, I'm not having much fun right now, but soon!
I actually look forward to work/don't dread Sundays at all.
PS: related HN article: https://news.ycombinator.com/item?id=31107128
When teaching ML, we mainly used pre-trained models. They learn the concept of data, train/test split, accuracy, and to build programs which use these models.
EDIT: in a man/years sense, we don't run pre-historic software. Although some of our clients are still on Win2008 and CentOS6...