Universal Fine-Tuning of ChatGPT through OpenAI’s Evals Framework: A Geo-Task Case Study

Discover how OpenAI’s “evals” framework enables targeted fine-tuning of ChatGPT across diverse domains. Using geospatial tasks such as automated map change as an example, this workshop will equip you with the skills to align ChatGPT for superior performance in any specialized task.

Hosts: Hannes Kruppa & Adrian Margin

Building and Scaling Large Language Model Applications: A Hands-On Workshop

Dive deep into the world of Large Language Models (LLMs) in this hands-on, interactive workshop designed to empower developers and tech enthusiasts alike. We will explore the cutting-edge applications and methodologies of LLMs, providing participants with the knowledge and tools to innovate in their own fields.

Session Breakdown:

Introduction to LLMs: Kickstart your journey with a guide to the current state of Large Language Model applications. Understand the different models available, delve into existing prompt engineering techniques such as chaining, agents, few-shot, RAG, etc. and familiarize yourself with the frameworks and services that underpin LLM applications.

Hands-On Development: Roll up your sleeves as we guide you through building a practical LLM application. Using the Open Food Facts database, we will create a versatile application capable of discovering food items, suggesting recipes, and planning shopping lists. This segment will solidify your theoretical knowledge.

Scaling and Pitfalls: Learn from our company’s real-world experience as we share the challenges and pitfalls encountered when scaling an LLM application to a mass product. Gain insights into effective strategies and approaches to mitigate these issues, ensuring your application is not just functional but also scalable.

Host: Andrej Romanov

Rapid Embedded Prototyping with Machine Learning

A technical presentation from Dyson’s New Product Innovation Software Team (NST) on tools and practices involved in developing high-tech prototypes using cutting-edge embedded systems technology, with a ‘hands on’ example application using Machine Learning.

Host: Marco Domingo

Function Calling Strategies: Conversations with Your Live Data

This workshop delves into leveraging Function Calling—a powerful feature of OpenAI models such as GPT-4 and GPT-3.5 Turbo—to facilitate natural conversations with our data. Participants will explore the immense potential of connecting GPT models with live data by programmatically interacting with them, gaining insights into triggering custom operations on our end. Throughout the session, attendees will understand how to integrate OpenAI models into data workflows and obtain data-driven insights. We will imagine a data store and show how we can obtain complex information by asking for it in plain English.

Host: Adrian Caliman