2025-11-05 –, AI
A hands-on guide to fine-tuning small language models (SLMs) using Python tools like transformers, unsloth, and trl. Learn practical workflows that empower data teams to build performant, private, and domain-specific LLMs.
This talk presents a practical guide to fine-tuning small language models (SLMs) using powerful open-source Python tools. While large language models (LLMs) are impressive, they often come with steep compute costs, limited domain adaptability, and privacy concerns. By contrast, SLMs offer a leaner, more customizable alternative—especially when fine-tuned on your data.
We’ll walk through a full SLM fine-tuning workflow using tools like transformers, unsloth, trl, peft, and accelerate, with a focus on practical examples and performance trade-offs. You’ll learn how to design lightweight pipelines, leverage LoRA adapters for cost efficiency, and deploy models optimized for your data and use case.
This talk is aimed at data scientists, ML engineers, and technical leaders looking to go beyond prompt engineering and harness the full power of Python for LLM customization. Attendees should be familiar with Python and basic ML/transformer concepts.
Previous knowledge expected
Sigal Shaked is a founder, technologist, and researcher with over 20 years of experience at the intersection of data, machine learning, and Generative AI. She holds a PhD in Software and Information Systems Engineering and was among the early academic contributors to the emerging field of GenAI. At Datomize, she led the development of a GenAI-powered synthetic data platform, and today, at Datawizz, she focuses on building domain-specific small language models (SLMs) that prioritize performance, privacy, and real-world utility.