YamFlow — Reference Machine Learning Workflow

YamFlow — Reference Machine Learning Workflow

Hello, World! We’re happy to announce our new startup YAM (www.yam.ai).

We initially chose YAM to stand for Yet Another Machine (which meant Artificial Intelligence) and later made it a recursive backronym YAM AI Machinery. At this startup, we strive to standardize the practices and frameworks of developing AI applications so that AI can be componentized for reuse and integration. With reusable and mashable AI components, we help enterprises build AI applications in a fast and proven fashion through consultancy.

YamFlow Draft Specification

First of all, we would like share with you the draft version of our reference workflow for the machine learning (ML) development life-cycle. We name this workflow YamFlow. YamFlow is aimed to provide a canonical taxonomy for practitioners to understand and communicate the flows of activities and data involved in a typical ML process. It specifies the key activities of pipelining data, modeling ML, training ML, and serving the inference.

Internally, we also use YamFlow as the baseline for YAM AI Machinery to design interoperable frameworks for composing ML tasks and data.

If you’re involved in developing ML applications in an enterprise environment, we believe you’ll find YamFlow useful too. While we are maintaining YamFlow as a live specification, we wish to hear your comments so that we can keep improving it for more accuracy, practicality, and generality.

Please check out YamFlow at https://flow.yam.ai, which is maintained as a GitHub project. We’d love to hear feedback from you so that we can improve YamFlow to cover your use cases.

Lastly, we’d love to stay in touch on social media:

• Website: www.yam.ai
• Blog: medium.com/yam-ai
• Github: github.com/yam-ai
• Twitter: twitter.com/theYAMai
• LinkedIn: linkedin.com/company/YAMai
• Facebook: facebook.com/theYAMai