When a child wants to ride a bike, they learn by doing. This process of trial and error is also known as learning through reinforcement, because positive and negative experiences promote or discourage certain behaviours, respectively. Children learn to ride by avoiding actions that cause them to crash; for the thrill of feeling wind in their hair.
Read more
This article shows how you can employ three frameworks to orchestrate a machine learning pipeline composed of an Extract, Transform, and Load step (ETL), and an ML training stage with comprehensive tracking of parameters, results and artifacts such as trained models. Furthermore, it shows how Pachyderm’s lineage integrates with an MLflow’s tracking server to provide artifact provenance.
Read moreWinder.AI helped CMPC, a large paper milling company, to optimise their production process by using reinforcement learning. CMPC are now able to automate industrial processes that were previously manual. This case study describes our approach and the results.
Read moreWinder.AI worked with Grid.AI (now Lightning AI) to investigate how Buildpacks can minimize the number of base containers required to run a modern platform. A summary of this work includes:
The video below presents this work.
Read moreWinder.AI worked with Protocol.AI to evaluate general-purpose computation frameworks as part of our MLOps consulting services. A summary of this work includes:
Winder.AI worked with Grid.AI to stress test managed Kubernetes services with the aim of reducing training time and cost. A summary of this work includes:
Artificial intelligence (AI) workloads are resource hogs. They are notorious for requiring expensive GPUs and large numbers of machines to train complex models on large datasets in a reasonable amount of time.
Read moreDr. Phil Winder shares experiences of Winder.AI’s MLOps consulting experience at a variety of large and small organizations.
In this talk he presents industry observations of MLOps team size and structure for a range of business sizes and domains. Learn more about how others structure their MLOps teams. Discover which problems you need to solve first.
Welcome to Winder.AI talks. A series of free interactive webinars hosted by Dr Phil Winder, CEO of Winder.AI, Author of O’Reilly’s Reinforcement Learning and one of the founders of the MLOps Community, covering a range of topics about the use of machine learning operations (MLOps), reinforcement learning (RL), and machine learning (ML) in industry today.
Read moreDr. Phil Winder shares experiences of Winder.AI’s MLOps consulting experience at a variety of large and small organizations.
In this talk he presents a white paper that discusses the differences between two leaders in the data engineering space – Databricks and Pachyderm. Learn how these two products differ, when to use each, and the pros and cons.
At the end of the talk Phil distils this information and presets best practices. There’s also some discussion of future trends and some ideas for aspiring open-source engineers.
Read moreDr. Phil Winder shares experiences of Winder.AI’s machine learning consulting experience at a variety of large and small organizations.
In this talk he focuses on packaging ML models for production serving. Learn about how the cloud vendors compare, what orchestration abstractions prefer, and how packaging tools seek to find the right abstractions.
At the end of the talk Phil distils this information and presets best practices. There’s also some discussion of future trends and some ideas for aspiring open-source engineers.
Read moreNot so long ago, developers used clunky consoles to provision infrastructure and applications. It wasn’t long before someone realized it was better to automate such a process via scripts and APIs. But it wasn’t until Hashicorp showed that APIs were not enough. Their insight was to declare a canonical representation of the infrastructure. You can then reconcile this declaration against the live view of the infrastructure.
In 2015-16 we helped WeaveWorks develop their cloud monitoring platform. We targeted Kubernetes but lamented the hackiness of our script-based deployment pipeline. The common pattern at that time was to run a script that ran kubectl apply over and over again.
