March 27, 2025 | 10:00 AM PT
We’re excited to release the first end-to-end platform for training and serving LLMs with Reinforcement Fine-Tuning (RFT). Discover how RFT is revolutionizing AI model customization with breakthrough capabilities that traditional supervised methods can't match. While others continue relying on massive labeled datasets, Predibase is first-to-market with enterprise-ready RFT that achieves remarkable results with as few as 10 labeled examples.
You'll Learn:
- Why RFT is a Game Changer:
See how RFT outperforms Supervised Fine-Tuning (SFT) in critical scenarios - dramatically improving accuracy on reasoning tasks and excelling even when labeled data is scarce or non-existent - Practical Implementation:
Watch a live demonstration of our intuitive end-to-end RFT workflow, from designing reward functions to deploying optimized models - Real-World Success Story:
Deep dive into our PyTorch-to-Triton use case where we taught an AI to convert high-level PyTorch code into optimized GPU kernels - without requiring a large dataset of matched examples. (We’ve open-sourced this model so you can use it yourself!) - When to Choose RFT vs SFT:
Gain a practical framework for selecting the optimal fine-tuning approach for your specific use cases and data constraints
Featured Speakers:
- Dev Rishi - CEO & Cofounder, Predibase
- Travis Addair - CTO & Cofounder, Predibase
- Arnav Garg - ML Engineering Team Lead, Predibase
Who should attend:
AI practitioners, ML engineers, technical leaders, and data scientists looking to maximize model performance with minimal data requirements.
Can't make it? Register anyway and we'll send you the recording.
