MLOps Meetup Apr 2025 (Two talks!)
Tonight we’ll be hearing from:
Danny Boland on “Scaling an LLM-Based Product Feature at Zoe". From Danny:
LLMs have enabled a wide variety of new product features and capabilities. While we often discuss the challenges and techniques involved in achieving a functional capability with an LLM, this is only the beginning of building a product feature at scale. This talk will cover some of the challenges in meeting non-functional requirements when using LLMs and the trade-offs involved when needing low latencies, scalability, manageable pricing and availability. We’ll explore these with an example from Zoe of using a multimodal LLM to recognise and evaluate meals via a phone camera, with requirements for low latency and high scale at peak times.
I am a lead engineer at Zoe, a science and nutrition company on a mission to improve the health of millions. I have over ten years of experience in building and deploying ML and AI at scale, having previously worked at Skyscanner, Bloom & Wild and Vodafone after a PhD at the University of Glasgow.
Csoban Balogh on "Designing MLOps for configurability and observability". From Csoban:
"The evolution of MLOps at Skyscanner has been driven by the increasing number of models we own and how to measure and manage these effectively. This talk covers how we use our Experimentation system and custom configurations such as Model Families to achieve this."
"I am Csoban (Cho - ban), a Graduate Software Engineer at Skyscanner, working on the Machine Learning Platform for enablement of ML within Skyscanner."
🍺 Beer and 🍕 Pizzas this evening come from our Sponsor, Skyscanner, jobs:
Software Engineer 2, Distributed Systems - AI Enablement (London or Edinburgh)
Senior Software Engineer - Distributed Systems - AI Enablement (London or Edinburgh)
We will also be hosted again in the Bayes Centre.
Plan for the evening:
18:30: doors open
18:30 -> 19:00: chat / pizza
19:00 -> 20:00: talks + q/a
20:00 -> 20:30: tidy-up / head to nearby bar
20:30: event close