Lin Qiao
CEO and Cofounder | Fireworks AI
Lin Qiao is the CEO and co-founder of inference platform Fireworks AI, which makes it simple, fast and cost-effective for enterprises like Uber, Quora and DoorDash to build and scale genAI products. Prior to founding Fireworks, Lin was the head of Meta’s PyTorch.
All Sessions by Lin Qiao
Compound AI Systems and the Future of AI Integration
<span class="etn-schedule-location"> <span class="firstfocus">Generative AI</span> <span class="firstfocus">AI Agents</span> <span class="secfocus">Intermediate</span> </span>The genAI landscape is experiencing a paradigm shift from single models to compound AI systems. These combine models across modalities (i.e., text, audio, video) with external knowledge sources into agentic workflows. By drawing on the unique strengths of different models, and providing access to realtime databases and knowledge sources, we can overcome the finite nature of the single-model approach and the probabilistic limitations of genAI. In this talk, Lin Qiao of Fireworks AI will discuss why compound AI systems are the future, the tools necessary to build them, and distinct design approaches and their tradeoffs. Single models have limited capabilities in problem solving, and app developers need access to compound AI systems to drive innovation with high quality. The presentation will dive into the design of a compound AI system that can automatically decompose a complex business task into multiple steps, accessing multiple models across many modalities (text, audio, image and more), retrievers, and external tools. Attendees will learn about the developer toolkit for fast iteration and deployment of compound AI systems. Key Takeaways: Understanding compound AI systems: Attendees will learn what compound AI systems are, and how they differ from single-model approaches. Designing and optimizing compound AI systems: Attendees will understand the best practices and strategies for designing and optimizing compound AI systems, as well as which platforms and tools to work with. Enterprise examples of compound AI systems: Attendees will see examples of enterprises building and deploying production-scale compound AI systems.
Compound AI Systems and the Future of AI Integration
<span class="etn-schedule-location"> <span class="firstfocus">Generative AI</span> <span class="firstfocus">AI Agents</span> <span class="secfocus">Intermediate</span> </span>The genAI landscape is experiencing a paradigm shift from single models to compound AI systems. These combine models across modalities (i.e., text, audio, video) with external knowledge sources into agentic workflows. By drawing on the unique strengths of different models, and providing access to realtime databases and knowledge sources, we can overcome the finite nature of the single-model approach and the probabilistic limitations of genAI. In this talk, Lin Qiao of Fireworks AI will discuss why compound AI systems are the future, the tools necessary to build them, and distinct design approaches and their tradeoffs. Single models have limited capabilities in problem solving, and app developers need access to compound AI systems to drive innovation with high quality. The presentation will dive into the design of a compound AI system that can automatically decompose a complex business task into multiple steps, accessing multiple models across many modalities (text, audio, image and more), retrievers, and external tools. Attendees will learn about the developer toolkit for fast iteration and deployment of compound AI systems. Key Takeaways: Understanding compound AI systems: Attendees will learn what compound AI systems are, and how they differ from single-model approaches. Designing and optimizing compound AI systems: Attendees will understand the best practices and strategies for designing and optimizing compound AI systems, as well as which platforms and tools to work with. Enterprise examples of compound AI systems: Attendees will see examples of enterprises building and deploying production-scale compound AI systems.
