Tonight we’ll be hearing from:

Assulan Nurkas on "Scaling Fine-Tuned LLMs: Processing 200k Daily Reviews at Trustpilot". From Assulan:

Moving from small language models to large ones is easy in a notebook, but deploying them cost-effectively at scale is a different challenge. In this talk, I will walk through how Trustpilot revamped its core Topic and Sentiment models using customized, fine-tuned LLMs. We will dive into the architecture of our reusable, enterprise-grade infrastructure designed to handle near-real-time processing of over 200,000 daily reviews. Attendees will get an inside look at the MLOps strategies we used to address latency, cost control, and quality evaluations. The result? A 55% reduction in model errors, significant cost savings, and a more sustainable, energy-efficient pipeline.

Assulan Nurkas is a Staff ML Engineer at Trustpilot focused on building robust ML infrastructure. With over 8 years of platform experience—including a background as a Senior Software Engineer at Craneware—he specializes in bridging the gap between research and production. Since 2021, Assulan has been dedicated to enabling Trustpilot’s data teams to deploy models at scale.

We will be hosted again this evening by Skyscanner, who are also sponsoring 🍕and 🍻. Many thanks!

Plan for the evening:

18:00: doors open
18:00 → 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

MLOps Meetup Feb