As part of MLOps chapter for 2022-23, our mission is to bring the MLOps community to better understand the science of managing Machine Learning (ML) lifecycle. We will focus on the science, decision-making process and approach based on experience and experiments that have led to success and war stories of implementing MLOps in organizations of various sizes and shapes.
MLOps is based on the similar principles of DevOps which talks about Continuous Delivery (CD) of ML models to production. We have seen MLOps gain traction in the past few years, and there have been tools and frameworks, including companies built around organizing and optimizing MLOps.
The topics which we will focus on as part of this chapter are:
- How do you know your MLOps process and culture is thriving in your company/team?
++ What are you measuring?
++ How do you quantify the impact of MLOps and validate the investment?
- How to design cost-effective processes, specifically inference landscape?
++ Cost of GPU inferencing vs CPU inferencing
++ Is your team investing in making the models leaner - distillation, quantization, pruning?
++ SLA requirements versus inference cost analysis.
- What are the common mistakes teams make with regards to implementing MLOps?
++ What were some of the mistakes you made and how did you resolve them over time?
- As an experienced MLOps practitioner what advice will you give to your beginner self?
- What are some of your favorite tools/frameworks for implementing MLOps?
++ How did they end up on your favourite list?
++ What makes them awesome?
++ How did you decide to buy, and not build?
Follow this project to:
- Track the latest updates on activities, content and opportunities.
- Comment on the topics and problem statements that MLOps @fifthel will be taking up.
- Register your interest in receiving updates and newsletter.
For more details, contact The Fifth Elephant at firstname.lastname@example.org or +91-7676332020.