Who decides the dirrection of Kubeflow
Key Takeaways
Kubeflow's direction is decided through a community-driven process, with a roadmap and versioning policy in place, led by the product manager and influenced by end-users and the community.
Full Transcript
and real fast on on medium you were writing about how new improvements keep coming out about every 90 days or so and as the product manager who is the one that's influencing where the product is going is it a team effort from everyone or there is there a vision that's written down how does that work that's a good question you know I should pay you for that yet yeah I've even seen this presentation and you're asking me they're good so along those lines we can make a process to be held to collect information from end users in a logical way because I think that's one of the reasons why we've been successful is we've tried to engage with customers and and and approach their their problems and improve their their workflows so one of the things that we do in coop lawanda which I was released on March 2nd I think even more important than the features that we delivered in one dodo was the process so we have a process to publish a roadmap right so people know what's coming and we do a roadmap by each individual components but also for kin flow in general we have a versioning policy so each component within Q flow will talk about how there are different components its composable and different components mature at different levels and different time frames we have stable beta and alpha and for stable components it's important that for us that we have an application requirements template and that template goes over the things that we would expect in that component to make it stable so this is like what the configuration and deployment requirements are and what CR DS custom resources are necessary and how do you log and monitor it what docker images are in there and what's the CI CD what's the documentation and testing and ownership and user adoption so I mean I think this is all important process that we've been able to deliver as important as the the code itself we're on our 9th release so it's not like it's a brand new release or a brand new project we track all this in a can be bored that everyone can see one dato was one of our largest releases with almost 200 issues that we've that we've delivered and really designed to improve the fit and finish and you can see here at the bottom right the components that we have we'll go through these in more detail that are in stable and in beta many of the beta components especially cute little pipelines will likely become stable in the 1.1 time frame 1.1 is scheduled for June of this year so I think that's that's a pretty good a review of some of the process
Original Description
What is the direction that the community of Kubeflow goes in and who is the one that decides it? For our 8th MLOps community meetup Josh Bottom VP of Arrikto and Kubeflow community product manager answers this question for us. This is taken from a longer conversation you can find here: https://youtu.be/jXRbj5xnBy4
Linkedin, Spotify, Volvo, JP Morgan, and many other market leaders are leveraging Kubeflow to simplify the creation and the efficient deployment of Machine Learning models on Kubernetes. This presentation will provide an update on the Kubeflow 1.0 release, and review the Community’s best practices to support Critical User Journeys, which optimize ML workflows.
As a data scientist will often need to build (and save) hundreds of variants of their model, this session will provide a deeper dive into how an integrated storage solution simplifies model-building and increases ML productivity. The presentation will examine how to optimize the daily workflows of data scientists, and eliminate complex and time-consuming manual tasks. The talk will also highlight how efficient Kubeflow operations rely on Kubernetes storage primitives, such as Dynamic Volume Provisioning, Persistent Volumes and StatefulSets. This integrated solution simplifies the configuration, operations and data protection for Kubeflow and generic K8s stateful apps in production-grade, multi-user environments.
In this chat we sit down with Josh Bottum, a Kubeflow Community Product Manager. His Community responsibilities include assisting users to quantify Kubeflow business value, develop critical user journeys (CUJs), triage incoming user issues, prioritize feature delivery, write release announcements and deliver Kubeflow presentations and demonstrations.
Mr. Bottum is also a VP of Arrikto. Arrikto simplifies storage operations for stateful Kubernetes applications by enabling efficient local storage architectures with data durability and portability. Arrikto is a core code contribu
Playlist
Uploads from MLOps.community · MLOps.community · 56 of 60
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
▶
57
58
59
60
Our 1st MLOps Meetup // Luke Marsden // MLOps Meetup #1
MLOps.community
Remote Collaboration as a Data Scientist
MLOps.community
MLOps Manifesto with Luke Marsden from Dotscience
MLOps.community
MLOps lifecycle description
MLOps.community
What Does Best in Class AI/ML Governance Look Like in Fin Services? // Charles Radclyffe // MLOps #2
MLOps.community
Life purpose and too many spreadsheets
MLOps.community
Explainability, Black boxes and EU white paper on reproducibility
MLOps.community
Hierarchy of Machine Learning Needs // Phil Winder // MLOps Meetup #3
MLOps.community
Automatically Retrain Machine Learning Models? Are best practices worth it?
MLOps.community
Building an MLOps Team? Key ideas to keep in mind
MLOps.community
Hierarchy of MLOps Needs
MLOps.community
Bare necessities for getting an ML model into production
MLOps.community
MLOps and Monitoring
MLOps.community
How Phil Winder got into Data Science and Software Engineering
MLOps.community
Provenance and Reproducibility in Machine Learning; what is it and why you need it?
MLOps.community
Friction Between Data Scientists and Software Engineers
MLOps.community
MLOps Problems in different size companies
MLOps.community
ML tooling in large companies
MLOps.community
ML Platforms - The build vs buy question
MLOps.community
ML Services Gateway at SurveyMonkey
MLOps.community
Message buses, Async and sync architecture
MLOps.community
MLOps #4: Shubhi Jain - Building an ML Platform @SurveyMonkey
MLOps.community
Hybrid Data Science Teams @SurveyMonkey
MLOps.community
How do you handle ML version control at SurveyMonkey
MLOps.community
Doing ML with Personal Information
MLOps.community
Evolution of the ML feature store @SurveyMonkey
MLOps.community
Developing a Machine Learning Feature Store
MLOps.community
Auto retrain ML models is not the question
MLOps.community
3 key parts to Machine Learning monitoring
MLOps.community
MLOps Meetup #6: Mid-Scale Production Feature Engineering with Dr. Venkata Pingali
MLOps.community
MLOps meetup #5 High Stakes ML: Active Failures, Latent Factors with Flavio Clesio
MLOps.community
MLOps: Airflow Pros and Cons
MLOps.community
Specific challenges in Machine Learning
MLOps.community
Current State Of Machine Learning
MLOps.community
Humans in the Loop are a defining factor in Machine Learning
MLOps.community
Learning from real life Machine Learning failures
MLOps.community
Survivorship Bias in machine learning tutorials
MLOps.community
Swiss Cheese model in Machine Learning
MLOps.community
Resume driven development in Machine learning & software engineering
MLOps.community
Who has the highest standards in ML?
MLOps.community
Venkata Pingali of Scribble Data Thoughts on the Current State of Machine Learning
MLOps.community
Dependable data and being able to Trust in your Data with Venkata Pengali of Scribble Data
MLOps.community
Speed, Trust, Evolution and Scale in MLOps
MLOps.community
More difficult transition for data scientists to become ML engineers
MLOps.community
How many models in prod til I need a dedicated ML platform?
MLOps.community
Deeper thinking from data scientists around platform blackholes
MLOps.community
Checkpointing, metadata, and confidence in your data
MLOps.community
Adjacent usecases and multistep feature engineering
MLOps.community
Standardization of Machine Learning tools like in Software Engineering with Venkata Pingali
MLOps.community
Reproducability flaws in end to end Machine Learning debugging
MLOps.community
3rd wave of data scientists
MLOps.community
MLOps meetup #7 Alex Spanos // TrueLayer 's MLOps Pipeline
MLOps.community
MLOps Meetup #8 Optimizing Your ML Workflow with Kubeflow 1.0
MLOps.community
Are Kubeflow and Airflow complementary?
MLOps.community
Why Kubeflow gained so much traction=open community
MLOps.community
Who decides the dirrection of Kubeflow
MLOps.community
What do Kubeflow and Arrikto do and how do they work together?
MLOps.community
Versioning your ML steps with Kubeflow
MLOps.community
Machine Learning Lifecycles//Perception vs Reality
MLOps.community
Kubeflow vs SageMaker in Machine Learning
MLOps.community
More on: AI Systems Design
View skill →
🎓
Tutor Explanation
DeepCamp AI