Meetup #2: Getting Down With Data

Hello Data & AI London! Thank you so much for joining us on our first meet up last month! We had an absolutely fantastic turnout last time and brilliant to see our community already growing so quickly. On that note...

We are super excited to welcome you all to our 2nd meet up hosted once again at our Mesh-AI HQ in Liverpool Street! We have hired out our 2nd floor space which is more open plan than our office. We also have updated and improved the AV equipment for the people at the back 👀 😊

Doors will be open from 6pm talks will start around 6.30pm. We will have food and both alcoholic and non alcoholic beverages provided throughout the evening along a fantastic line up of talks..

Talk #1 : Machine Learning on Kubernetes
Salman Iqbal | Appvia

In order to run Machine Learning (ML) models at scale three things are required: code, ML framework of your choice and a platform to run your code on. A number of ML frameworks exist such as TensorFlow, PyTorch, MXNet, Keras etc. When it comes to platforms, many of those exist too. Kubernetes is one such platform that runs containerised workloads. Kubernetes provides features that solve some of the challenges faced by data scientists.
This talk will discuss how it helps with infrastructure management, auto scaling, auto recovery from failure, automated deployments, multi-tenancy and GPU offloading. This talk will also show how to run some of the ML Frameworks on Kubernetes such as Scikit-learn, PyTorch, MXNet & TensorFlow. The talk will also discuss when not to use Kubernetes to run your ML workloads.

Bio
Salman works as an MLOps Engineer at Appvia and a Kuberenetes Instructor at Learnk8s. He has worked with a number of organisations in setting up Machine Learning platforms for teams to operate at scale. You can also find him on YouTube as Soulman Iqbal where he tries to explain cloud native concepts by simplifying them.

Talk #2 : Lessons Learned on building ML Platforms
Raul Ferreira | Mesh-AI

The increasing adoption of ML by organisations is starting to surface bottlenecks to innovation due to unnecessary manual workflows and lack of governance around ML assets. A reference MLOps architecture supports a clear operating model by defining how different users, such as data scientists, data engineers, ML engineers, IT, and business stakeholders, should collaborate and interact. We'll discuss lessons learned around the different challenges that data science practitioners face across the different phases of the machine learning lifecycle, including experimentation, productionisation and monitoring. We will also cover the set of frameworks and best practices that can be used to mitigate these challenges.

Bio
Raul is Principal Consultant at Mesh-AI. A full stack ML developer with 10 years of experience. He has previously built ML platforms to serve dozens of researchers hosting over +1k models in production with complex dependency chains. Experienced in designing tools and processes to significantly reduce lead time to bring world class Machine Learning research to production.