2025-09-20 –, Ladd Room (Capacity 96)
In this talk, we will introduce SDG Hub, an open-source toolkit developed at Red Hat for customizing language models using synthetic data. We will begin by unpacking what synthetic data means in the context of LLMs, and how it enables model customization.
The session will explore SDG Hub’s core components: prompts, blocks, and flows, and demonstrate how users can compose, extend, or modify pipelines to fit specific tasks. It will also cover strategies for choosing the right teacher model depending on the use case (reasoning, translation, etc.), and walk through two real-world examples: building a document-grounded skill using a pre-built pipeline, and customizing a reasoning model by authoring new blocks, prompts, flows, and integrating a custom teacher.
The talk will conclude with a demo of the new SDG Hub GUI, showcasing how non-experts can visually construct and manage their own synthetic data pipelines.
Advanced - background knowledge and skills are required
Shivchander Sudalairaj is a Senior Research Engineer at Red Hat AI Innovation Team, where he leads the development of tooling for synthetic data generation to customize large language models. His work focuses on customizing large language models for enterprise use through instruction tuning and inference-time scaling. Previously, at IBM Research, he contributed to LLM alignment for the Granite model family and methods for generating differentially private synthetic data. He holds a Master's in Computer Science from the University of Cincinnati.
Abhishek Bhandwaldar is a Research Software Engineer at Red Hat and IBM Research, leading knowledge infusion efforts for RHEL AI. At the MIT-IBM AI Lab, he worked on reinforcement learning, language model alignment, and synthetic data generation to improve retrieval-augmented generation and code models. He holds a Master’s in Computer Science from UNC Charlotte.