Ask HN: How to break into AR/MR Industry | Hacker News Reader

Ask HN: How to break into AR/MR Industry

2 pointsmendeza7 comments

I am in graduate school at Cornell and I am super passionate about augmented reality and mixed reality. I am in need of advice on how to break into the AR/MR/VR industry. I am having a hard time getting interviews for top companies (Magic Leap, Oculus,Apple), and it is incredibly frustrating since I feel I am very qualified to work at the companies in the role of a computer vision or graphics developer.

I have strong knowledge in computer vision and computer graphics, and I have worked on projects at MIT and Verizon building Augmented Reality apps. I keep having a hard time getting interviews for top AR/VR companies such as Magic Leap and Oculus, and I really want to work in these companies to be at the forefront of the industry.

I am thinking starting to freelance for small companies and brands to build AR as several companies have reached out to me to develop AR, but I feel that working for a company that is pushing the industry forward will be more valuable.

Any advice will be super helpful, and I am interested in hearing about any opportunities if anyone knows of anything. Shoot me an email at hub503bub@gmail.com.

Oh! After this much education, u have bias. The only MR working gadget, Hololens is from MS. Did you try them? They are ahead of every one.

How do I have bias? I have seen many different platforms facilitate AR and how AR has progressed since 2011. How do you even define MR, when Google coins MR for their VR head tracking, but magic leap talk about MR as being able to see in front of you.

You are mistaken that hololens is the only device that does MR. Occipital does better MR in my opinion, as they have a larger FOV and they leverage SceneKit, an amazing platform to design 3D games and software. Have you seen Daquri and Meta's MR headsets as well?

I'm not in the current AR/MR/VR industry, but I have friends and colleagues who are, and I worked in VR last time it was a thing.

I'd guess you're missing one or more of: (a) serious EE/ME experience, (b) having contributed to computer vision/graphics instead of just understanding the current state of the art and/or tooling, and/or (c) having designed and built AR/MR/VR hardware and software systems from scratch instead of building apps for existing systems or using existing tooling.

At a company that is inventing something new, the most valuable, sought-after hires are of people who invent and advance the company's work; and people who can interface between engineering and content.

The former means people who invent and write algorithms for OpenCV, not people who simply use OpenCV. Or having built AR hardware and software in combination, and therefore have a deep, practical understanding of the tradeoffs in different generations of hardware and different methods of projection, etc., rather than having only built apps for existing AR toolkits or platforms.

The latter means software developers who can also work with the hardware. They may not be required to design a circuit board or engineer some optics, but they understand the operation of the hardware, can fix it if it breaks, can diagnose hardware problems and provide the sort of hardware bug reports every software developer wishes an end-user could provide, etc.

(An example of this would be the technical artist in video game studios. They're a respected, senior artist [so they can make actual art, and so their processes to do so are reasonable and representative] who can also program, and either creates tools for artists, or is the interface between the artists and the tool developers.)

If you aren't one or the other, then you're a step removed from where the company is at: inventing new things. They don't have the time or manpower to hold your hand. You're stuck only being able to build for the things that already exist, not helping them advance.

It's why Microsoft's #1 recommendation for non-technical designers who want to get into HoloLens work is to become technical enough to be proficient in Unity, if not the entire Windows 10 development toolchain. If they aren't self-sufficient enough to be able to get all of their work onto a device on their own, HoloLens isn't ready for them yet. It's not plug-and-play, everyone involved needs to be able to wear all the hats.

If you can't show that you're already inventing new things yourself; or that you have a practical, working knowledge of the entire system, the kind that comes with having built something from scratch; then there may not be a place for you yet.

Yes. Vito is correct. There are 2 kinds of employees. Repeaters - Fit for operations, repeat the operations, increment the operations and using the tools.

Innovators - Sharp,Imagine tools, Tools used innovatively and think out of box, not afraid of being called as 1 with crazy ideas and innovative.

Thanks for the feedback vitovito! That was my reasoning for me to enroll into grad school, so I can go beyond understanding the tools and build cool apps, but to push the state of the field. I feel I am at point in my knowledge where I can contribute the field of AR/VR, I just need guidance on what problem to focus on.

The biggest thing that I would want to know is what software/algorithms/knowledge would advance the work of an AR/VR company. From these job postings:

1. Apple: 3D Perception/Computer Vision Algorithm Engineer 2. Snap Inc : https://boards.greenhouse.io/snapchat/jobs/586658#.WLNBhBIrK... 3. Magic Leap: https://www.magicleap.com/#/job-post/521175

That most AR/VR companies are looking for people who are doing real-time computer vision algorithms, such as 3D object recognition, SLAM, and gesture recognition. I feel that this would be something I can focus on, but I do not want to work on a problem that is already solved. Any advice on what are important problems that would be of value to solve would be great!

Would implementing these algorithms and open them to the OpenCV community be of value also? I am only in a 2 year masters program, so I am limited on time to what I can contribute as a full-time grad student.

Any other advice on this you feel is important would be greatly appreciated!

I have no idea, I'm not either of those types of people for the current VR/AR/MR industry. And, even if I were, whatever I might say is not certain to be useful to you.

Normally, I'd recommend you interview people working in the space to understand what problems they're encountering now, and what they expect to run up against next, but when they're inventing as they go, by definition, they don't (and can't!) know.

They could tell you their current problem(s), but neither you nor they know if that's a general-case problem, or a problem specific to their implementation. Plus, they'll solve it in parallel before you, because they have the compounded knowledge of all their work leading up to it.

Only the people actually working on the problems know what those problems are, and they don't have time to stop and bring you up to speed. You have to catch up, or at least show you're on the way on your own.

The job listings seem to support this:

Apple wants either a PhD, which means having published papers advancing the field; or industry experience delivering products, which means having discovered problems and then solved them for real users and getting paid for them, end-to-end.

Snap wants a Master or PhD, or "experience implementing, verifying, and optimizing," which is just short of being responsible for shipping a product.

Magic Leap is an internship, but they want "Demonstrated experience," aka having built something yourself.

In this, you can't be a generalist, up for solving any problem. I'm sure you're entirely capable of that, but the requirement is also that you can figure out what the problems are and will be, on your own. In a new industry, there aren't paths to follow, everyone's still cutting down brush.

They're not publishing most of their work, so you have no idea what's solved, and even if they were, I think it's generally accepted that a published paper (or launched product) is a few years behind what's currently being tried in a lab.

You will probably inevitably be redoing someone else's work, but you won't learn that for months or years.

I'm not sure there's any advice I could give except to pick a problem domain, pick a use case, learn all the component pieces of that problem, make new and different tradeoffs, and figure out what's unsolved for the choices you made.

Reply on news.ycombinator.com