The “Chicken or the Egg” paradox is a recurring problem in modern software development. In so many words: if you want resources allocated to build something, you better have already built it - and have the numbers to show that it is worth it.

Such predicament is increasingly relevant for machine learning, specifically when the fruits of R&D are ripe enough for “production.” The entire ecosphere of MLOps has evolved around technical debt and related problems that obstruct productivity, but solving such problems in various settings requires an inter-organizational value proposition (chicken), which in turn requires some evidence (egg). 

A traditional solution is to “build” internal implementations which optimize for integration ease. Most of these “interim” solutions take a life on their own and will only be replaced once (if) they outlive their usefulness or another abstraction is adopted. As such, they represent the most cost-effective approach for an organization. 

This webinar introduces “Extensible MLOps,” enabling an internal solution with solid open-source foundations and a “lean” stack. Essentially, it is a familiar solution for a familiar problem. We will claim that like in other realms of software development, any “Chicken or the Egg” paradox in MLOps can be solved, specifically, by:

  • Involving all team members in the initial design of the workflow

  • Rolling a solution with low integration costs (in terms of lines-of-code)

  • Iteratively adding MLOps functionality onto existing R&D codebases

  • Producing viable pipelines (still in R&D)

  • Running a POC and collecting improvement metrics

Specific code examples and architectures will be discussed, relying on the open-source ClearML MLOps solution. An extended Q&A session will be held following the presentation.

Instructor's Bio

Ariel Biller

Evangelist at ClearML

Researcher first, developer second. Over the last 5 years, Ariel has worked on various projects; from the realms of quantum chemistry, massively-parallel supercomputing to deep-learning computer-vision. With AllegroAi, he helped build an open-source R&D platform (Allegro Trains), and later went on to lead a data-first transition for a revolutionary nanochemistry startup (StoreDot). Answering his calling to spread the word, he recently took up the mantle of Evangelist at ClearML. Ariel received his PhD in Chemistry in 2014 from the Weizmann Institute of Science. Ariel recently made the transition to the bustling startup scene of Tel-Aviv, and to cutting-edge Deep Learning research.


  • 1

    ON-DEMAND WEBINAR: Embracing Extensible MLOps - Solving “Chicken or the Egg”

    • Ai+ Training

    • Webinar recording

    • Join ODSC West 2021 Training Conference