Meeting link: https://meet.google.com/vtk-snhb-fjc
Weekly meet time: Thursday 5-6pm EST
We coordinate using #dg-mlops-fun on the AISC Slack.
We’ve been meeting since Jan 2021 to cover a number of topics. For this cycle we’ll focus on planning the infrastructure of a chat bot service.
We’ll dive into application architecture, data system design, infrastructure design and using a simple open source model.
Brainstorming and Gathering requirements
Designing an architecture v1
Investigating training vs inference architectures
Dealing with data drift and retraining
Deploying a sample model (Rasa)
Create recipes for a sweet afterparty!
We’ll be working on recipes as part of the working group to capture some of the knowledge we created.
For our final meeting we’ll have a social and you’ll get a gift card to buy something based on your contributions!
Week of March 10 - Intro and Brainstorming
Week of March 17 - Intro to Architecture
During Week two we’ll be starting thinking about designing a logical architecture. We’ll also introduce recipes and how we can contribute to them.
- Diagram per group of a potential logical architecture
- How to create a logical architecture recipes
Sample architecture (Rasa)
Designing a E-commerce Data Platform
Jan 27 MLOPS Requirements Doc
Based on your research, work with a group to think of some *reasonable* requirements based on the non functional requirements. Mr E is a business person, but is leaning on how you to advice on the technical parts of it. Brainstorm some requirements that you will present to Mr E for approval.
ML Ops Security
Through the spring of 2021, we focused our discussions on ML Ops Security, from these I wrote 3 articles to summarize some of the topics.
7 Layers of MLOps Security
So you've heard of MLOps and you've deployed a model to production! Awesome, but now that you've hit your milestone, you should probably double check to make sure that everything is nice and secure. Security is a never ending topic but implementing the foundational aspects will help you sleep better at night.
Protecting Your Data Engineering and MLOps Storage
Welcome to Part 2 of our MLOps security guide! To recap, last time we discussed how to protect data and went through an example for a restaurant review app. We'll continue to reference concepts from that article and build on them as seen in the Table of Contents.
14 Principles To Secure Your Data Pipelines
Welcome to Part 3 of our MLOps security guide! To recap, last time we discussed how to protect your data storage and went through a comprehensive list of storage solutions. Today we'll dive deeper into understanding and protecting our data pipelines. Part 1. Intro and Data Security Part 2.
In our initial we discussed how different parts of the ML Ops fit together and what technologies exist. We asked people to come up with mindmaps, recipes and architectures along the way. A couple examples included below.
What Is MLOps? - RECIPE | Aggregate Intellect
Objective: For me, a successful ML app is not only about the accuracy of the Model, but also the infrastructure around the model to make the model a scalable, stable application.