7 MLOps breakout sessions I’m looking forward to at Re:Invent 2021

Rustem Feyzkhanov
3 min readNov 16, 2021


MLOps is an emerging field of best practices for businesses to run Machine Learning workflows in production. MLOps captures a very wide range of tasks from optimizing model inference in production to structuring training processes and data pipelines. New tools and practices appear every year so here are the top 7 AWS Re:Invent 2021 sessions to help you stay in the loop of MLOps in the AWS cloud. Also, check my blog post about Re:Invent 2020 here.

This year Re:Invent will have both in-person and virtual experiences this year while also celebrating 10 years. The virtual experience is free and you can register for virtual attendance at the AWS Portal with access to keynotes, leadership, and breakout sessions.

Here are the 7 MLOps breakout sessions which I’m looking forward to. All these sessions will be available both in person and online on Re:Invent platform.

Sessions about MLOps best practices.

MLOps is a very hands-on field so the best way to learn is by learning how other companies solved the same challenges and checking if they are applicable to your case.

Implementing MLOps practices with Amazon SageMaker (AIM320)

This session will cover MLOps in breadth from setting up a research environment to monitoring models in production using SageMaker. Vanguard will share how they are using SageMaker to train and deploy models at scale.

MLOps at Amazon: How to productionize ML workloads at scale (AMZ302)

One of the main challenges in a big company is how to make ML solutions reusable between different projects. Amazon will cover how they reduce production-level ML infrastructure delivery time from weeks to hours.

AI/ML for sustainability innovation: Insight at the edge (AIM207)

Deploying model on the edge device is a more complex process compared to cloud inference which includes additional challenges. AWS will share how they did it for the largest-scale fisheries on the planet.

Sessions about model deployment optimization.

Organizing model deployment in production is a challenging process. One needs to take into account speed, scalability, and cost to make sure that the model adheres to production requirements. There is no one fits all solution in the field so the approach always depends on the context. These sessions will give you an idea what are the current ways of deploying models to production and how to compare and benchmark them.

Automatically scale Amazon SageMaker endpoints for inference (AIM328)

This session will cover how to use Amazon SageMaker for building scalable inference to make sure that you don’t have large upfront costs while also being able to handle peak loads.

Achieve high performance and cost-effective model deployment (AIM408)

This session will cover Amazon SageMaker inference features in breadth from multi-model endpoint to CI/CD integrations. The session will help to choose a better inference option for your ML use case.

Accelerate innovation with AWS ML infrastructure services (CMP314)

Finding the right instance type for inference is a challenging task. For example, GPU does provide faster inference, but CPU will have cheaper idle time and will be more scalable. This session will cover how different instance types could be used for inference, what are their benchmarks and how they could be used for your use case.

Easily deploy models for the best performance & cost using Amazon SageMaker (AIM417)

Since there is no one-fits-all solution for model inference, the best option is actually to try out different options and choose the best based on the benchmark. This session will cover how to do so with Amazon SageMaker and find the right instance types and model optimizations.



Rustem Feyzkhanov

I'm a staff machine learning engineer at Instrumental, where I work on analytical models for the manufacturing industry, and AWS Machine Learning Hero.