Explore a new approach to building complex workflows with Large Language Models (LLMs) in this 51-minute podcast episode featuring Vikram Rangnekar, Software Engineer at Stealth. Dive into the world of Ax, a TypeScript library based on Stanford's DSP paper research, which introduces concepts like prompt signatures, prompt tuning, and composable prompts. Learn how these innovations facilitate the creation and maintenance of RAG and agent-powered ideas previously challenging to implement. Discover insights on fine-tuned models, JavaScript tool evolution, DSP knowledge distillation, and optimizing task contexts. Gain understanding of API type validation, LLM value and innovation, navigating complex systems, and exploring LLM personas. Uncover strategies for optimizing small agents and tackling complex tasks with LLM assistance.
Ax: A New Way to Build Complex Workflows with LLMs - MLOps Podcast
Overview
Syllabus
[] Vikram preferred coffee
[] Takeaways
[] Data Engineering for AI/ML Conference Ad
[] Vikram's work these days
[] Fine-tuned Model insights
[] Java Script tool evolution
[] DSP knowledge distillation
[] DSP vs Manual examples
[] Optimizing task context
[] API type validation explained
[] LLM value and innovation
[] Navigating complex systems
[] DSP code generators explained
[] Exploring LLM personas
[] Optimizing small agents
[] Complex task assistance
[] Wrap up
Taught by
MLOps.community
