There are usually some trivial tasks around docs or simple items that can be done with minimal knowledge of the language or project codebase, additionally there are more low-mid tier tasks that require looking at past commits / structure of the project but follow the same pattern.
source: I have contributed to probably close to 100 OSS projects in the last decade varying in size and complexity.
Specifically With respect to kubernetes I have spent some time with trivial tasks for related projects like acs-engine because I am using azure for deployments.
I have not contributed to TensorFlow but in limited areas I imagine I could help with bindings, cross platform support and maybe optimisations of functions given that this is important because the compute cost involved. I would not however be able to dig into some of the complex models the codebase because I have not studied data-science or in any great detail.
We are in an era where OSS project organisers can now see the value and build on these small changes. If you can get enough people interested and contributing towards a goal you can reach a critical mass.
The one tip I would have if you were looking to get started out in OSS development is to make some kind of assessment around how easy it would be to work on a project.
For example: Rust is very popular right now and has been gaining traction for awhile. This is no accident and building a programming language out in the open is very very difficult. A number of very skilled Engineers have allocated time to mentoring. So if you wanted to just jump in having never contributed before you could just filter an issue by mentor. https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Ais... Filtering this further by E-easy to get yourself started.
This is probably not the best example as writing a programming language is not an easy feat. The example was chosen however to more highlight what to look for in projects if you want to get your feet wet and start contributing to OSS.