If you have a deep learning project, this is your to-go tool
What do you like best?
- Valohai allows easy management for deep learning, which is usually covered by a multitude of tools and is a hassle to manage. It brings all the tools you use in one place and therefore, besides huge amounts of data that your machine learning algorithms have to deal with, you don't have to deal with several various platforms.
- Version control for machine learning algorithms. I think this is one of the major value-added points of Valohai. Most of the time, you get only the beginning set of data and the "learned" result (with no idea or a basic idea of what happened in between). Valohai allows you to track all that info and therefore give you an option to repeat the experiment changing a couple of factors that didn't point your algorithm in the right direction (instead of guessing and trying)
- pipeline automation is another feature of Valohai's platform which promotes API-first development, therefore it's easy to integrate the pipeline into your existing development processes
- easy to scale up the project
- Valohai has good example cases that prove that this thing works
What do you dislike?
So far, I haven't encountered any problems. It supports a multitude of various ML/AI/Deep Learning/Other tools, therefore it adapts to what I do, not the other way around.
Recommendations to others considering the product:
It takes care of many headaches connected to the machine learning (like integrating with various tools you already use and don't want to switch, or scalability, or the version control which helps to pinpoint the moment when something went wrong - or right.)
What problems are you solving with the product? What benefits have you realized?
The company I work for creates software products with machine learning features and so far it was the best tool we have used for deep learning management.
