Exploring LLMs: Strategy & Practice

We're super excited to be back for our second meetup, and first meetup of the year! In this meetup we'll be exploring the topic of large language models (LLMs) and hearing from speakers around how to get value from LLMs effectively -- both from a strategic perspective and also from the applied perspective of a team that has run the gauntlet of using LLMs at scale to for content generation, learning some valuable insights along the way.

Here are our two presentations:

Effective LLMs: Drawing the rest of the owl, Lilly Ryan & Ned Letcher

Everyone is scrambling to embed LLMs into everything from bank chatbots to business strategy generation. For better or worse, they are being used as therapists, editors, educators, and financial advisors. But how long can these applications run before something goes entirely off the rails? Why do some use cases fare better, for longer, than others? And why are we so eager to accept the significant risks of trusting auto-generated content?

Lilly and Ned draw on their experiences in natural language processing, cybersecurity, philosophy, history, and consulting to offer advice on how to make the best use of LLMs, how to predict where they are likely to cause embarrassing failure, and the surefire ways to tell the difference. We’ll cover strategies for how to think about incorporating LLMs into a project, how to choose the right tool for the job, and why this matters, even when it feels like LLMs can do everything.

Whether you’re just starting out with generative AI, are a seasoned ML engineer, or you’re wondering how to make good software when you’re asked to incorporate non-deterministic chaos into your ordered ecosystem, this talk will give you the mental models you need to enter your next project with a solid strategy for minimising bullshit and getting the best outcomes—with or without LLMs—from beginning to end.

Lilly Ryan is a lapsed historian, information security architect, and catastrophiser-in-residence at Thoughtworks Australia, with a decade of experience in the fields of software development and web application security. Her humanities background, cybersecurity career, and stubbornness have led to a robust sideline in jailbreaking LLMs in the wild. In her "spare time" she serves on the board of Digital Rights Watch, co-hosts Byte Into IT on Melbourne radio station 3RRR, and wrangles Python conferences.

Ned Letcher is a lead data science engineer at Thoughtworks Australia. He’s worked across a range of sectors and domains, applying supervised and unsupervised machine learning, natural language processing, and data visualisation to business challenges and opportunities. Ned has used these experiences to develop strategies for making effective use of data & AI for identifying and framing the business value of data science initiatives.

LLMs for SEO Content Creation (in a non-evil way), James Bardsley

This presentation will detail the journey James and his team went through, using an LLM to generate multi-language content at massive scale for a global travel website. The triumphs, mistakes, and everything in between.

James Bardsley is a software engineer, product lead, team builder and curious poker of interesting things. James’ passion is solving problems, especially when the solutions help people connect with and understand their data. Talk to James about data visualisation, the role of automation in digital marketing, or your favourite beer!