2022 was an awesome year for MLOps and the MLOps London meetup. Many more companies realized their need to automate their ML life-cycle, techniques and best practices started to become standardized, and the data centric movement continued to gather steam, even making its way into the deployment space.
It was also a great year for our MLOps London community. We gained almost 1000 new members and hosted 6 incredible in-person meetups, all of which are available on demand on the YouTube channel.
A Masterclass on distributed training
If our January MLOps London meetup was anything to go by, 2023 is going to be even bigger and better. We started the year with two fantastic talks from Uroš Lipovšek and Paul Hetherington. Uroš kicked us off by talking about the complex world of distributed training. He highlighted how you can use tools like Kubernetes and Slurm to achieve a distributed training system. Then, he taught us some clever techniques for maximizing the GPU usage during those forward and backward passes.
You can catch the full replay of his talk here.
A Pythonic approach to GPU Inference
Having had a masterclass about training on GPUs, we got to complete the set by watching Paul Hetherington talk about GPU inference.
Paul took us through the work they’re doing at Mystic and how, using a pythonic approach, they deploy your inference code directly to GPUs. He also covered how GPU workload scheduling works and how they’ve been able to optimize it compared to traditional approaches.
You can watch the full recording of Paul’s talk here.
Networking and the future of MLOps London
As always, the evening was capped off with plenty of networking time. Lots of excellent conversations were had over a drink or two! If you’re local to London, make sure you come along in person to the next meetup which takes place on 28th March. If not, you can follow all the action live on our YouTube channel.