What sorts of side projects would be good for ML infrastructure?
I appreciate the value of side projects mentioned on Taro, and I'd like give one a shot. Whenever I've looked before, I got caught up in not finding ones relevant to my work. I work on the backend side, working in the machine learning infrastructure/ops space. That's literally the example used in the side projects video of one where it's hard to do on your own... so is it still worth doing a side project? Maybe one in a similar language (C++), even if it's not related?
I feel like I'm just not being creative enough here, but I'd rather not do a side project that's super unrelated to my area (e.g. learning a ton of frontend to build an app that has a small machine learning component) unless it's still valuable. Maybe to make this more general, how did those of you working on the backend get started with side projects?
Discussion
(3 comments)- 3a year agoRobinhood, Meta, Course Hero, PayPal
I Googled for "machine learning side projects" and found this article which seems pretty good.
Making a delightful, easy-to-use front-end is much harder than people think, so I agree that it's probably not best to pick it up just for a side project (unless you're considering a pivot into that area). The back-end projects I've seen be successful are generally REST APIs. To expose it, you can bundle whatever service you make with a Postman collection (just include it in the GitHub and add instructions on using it to the README).
- 3a year agoEntry-Level Software Engineer [IC1] [OP]Nvidia
I may have been getting too caught up in doing something super specific to my area (MLops or infrastructure). I haven't used too much REST or any Postman, so I'll look into both of those. Thanks, Alex!
- 2a year agoMachine Learning Engineer
As an aside, another way to determine side projects in MLOps is by understanding what's out there. You can know this by reading what other companies are doing with infrastructure. For example,
- You can see how netflix schedules their jupyter notebook jobs using papermill. Here is the blog too.
- You can see how uber uses a feature store to manage features ML Engineers pass into machine learning models. Here is a link to open source versions of feature stores.
- The MLOps community has great resources too.
Not saying you should implement everything from scratch. But reading does help to see "What kind of work would I be associated with as an MLOps engineer".
