GenAI Master Class: From Prototypes To Production Series – Pt 2: Evaluate and Productionize AI Systems

​Every company sees the potential of GenAI, but most teams hit a wall when turning it into scalable, production-ready applications. This series of master classes will equip you with the practical skills to build and deploy GenAI solutions that deliver real impact.

Schedule

• ​Nov 4: Part 1: Foundations of Prompting, RAG systems, Agents

• ​Nov 26: Part 2: Evaluate and Productionize AI Systems

• ​Dec 16: Part 3: Build and deploy a GenAI app end-to-end

​In this immersive program, you’ll:

• ​Build a GenAI app in class: Get hands-on experience designing and implementing a working GenAI solution from start to finish.

• ​Master RAG: Learn how to integrate retrieval with generation to build AI systems that dynamically pull from vast knowledge bases, delivering more accurate, contextually relevant, and up-to-date responses. In this section, you'll explore retrieval strategies, optimize how your models fetch and generate data, and fine-tune the balance between retrieval and generation for maximum performance in real-world applications.

• ​Build intelligent agents: Develop autonomous systems that don’t just generate outputs but actively make decisions, adapt to changing data, and perform tasks in real-time. These agents can automate complex workflows, handle unpredictable environments, and continuously learn, allowing you to build smarter, more responsive applications that drive efficiency and innovation in your business.

• ​Scale and Productionize AI Systems: Master techniques for building scalable, secure, and robust AI systems. Covering topics like red teaming, hallucination prevention, guardrails, evaluations, performance tuning, and data pipelines at scale.

​Who Should Attend:

• ​Technical Teams: Deep dive into GenAI architectures, RAG, and the latest AI frameworks.

• ​Technical Executives: Learn to identify high-impact GenAI use cases and build a roadmap for integrating AI to drive innovation.

​Part 2
In part 2, Charles Frye of Modal will cover what it takes to run production-grade LLM inference with vLLM. In addition to a high-level overview of the library, we will discuss reasons why you should (and when you should not) choose to build on vLLM. We'll also consider key optimizations for specific LLM inference workloads.

​We will also discuss evaluation best practices, because models are temporary and evaluations are forever.