For self driving car companies like Cruise, is it worth joining non robotics roles like data or ML infra? How is the work different from comparable roles at other internet companies? I don’t have relevant experience in robotics or computer vision and I would like to eventually get into that space.
TOP 80 Comments
- Cruise Automation UHDd18The work and challenges are far more rewarding than anything you’ll find at any internet company. You don’t necessarily need to be interested in robotics or computer vision to join the teams you mentioned - but it’s definitely the best way to get your foot in the door if that’s what you want to pursue
- The challenges on ML infra side are amazing. Definitely more satisfying when you see the results after so much grind.
Other roles that have massive impact are building great tools that would help robotics/ML folks visualize their algorithms, or contribute to software for the simulations
- There are a lot of design challenges in infra.
You should listen to Andrej's talk on Software 2.0 . I don't want to give out specific examples at the fear of talking something i should not, but that talk gives a good idea on how important ML infra is.
(I don't work in infra, so my comment isn't biased).
You also have non-ML side of infra which is piping a lot of data to engineers in a way that can be useful to them. Same with tooling. I know a lot of people who love having other engineers as their customers and feel satisfied when their work drastically improves the work flow.
It may not be math which people claim to be cool. But the design challenges are insane. I find a lot of them harder than the algorithms i deal with daily.Mar 16 1
- NIO / Eng Bye!Bye!OP, thanks for not categorizing Computer Vision under the Deep Learning bucket.
As for transitioning to ML and eventually CV/Robotics - once a software engineer, never an algorithm engineer. And vice versa, unless you undergo a Copernican inversion in world outlook.
- Self driving was hottest 5yrs ago now it’s saturated and has already been solved. What’s holding it back from going farther than a car feature are ethical issues and trust factors. Source: me, a previous CV engineer on an autonomous car project.
- Wait! Ca someone help me understand how ML/vision people would be less important then robotics people? I mean robotics people's work is useless without ML and vice versa so how come one could be better than the other. Other than not being able to finding people experienced in the area. I mostly look at it in one's value in the work place. Not necessarily in terms of the money.
- Microsoft interestiI think ML here is not what you think. By computer vision, it means pattern recognition etc that is needed to analyze and make a decision in real time which uses ML. What OP asks is a platform where you can cook data to analyze to improve the algorithm afterwards based on the data collection. The first part requires someone trained in ML for specific scenarios. The latter part is software engineers building a platform for data scientists and the vision scientists mentioned in the first part.