LLM Developer Day

Learn practical methods for designing and implementing LLM-powered systems on real-world business data using popular, ready-to-go LLM APIs—no specialized hardware, model training, or tricky deployment required. We'll show techniques for engineering effective inputs to the models (“prompts”) and how to combine LLMs with other systems, including business databases, with toolkits like LangChain. Join us and learn how to build LLM systems to generate tangible business results.

Push LLMs beyond the quality limits of off-the-shelf models and APIs by customizing them for domain-specific applications. We'll discuss strategies for preparing datasets and showcase gains from different forms of customization using practical, real-world examples. Join us and learn about model tuning techniques applicable for both API-based and self-managed LLMs.

*Please choose either Session 2a or Session 2b.

In this session, we'll explore foundational AI models in biology, as well as practical protein engineering and design applications supported by real-world examples. We'll discuss recent biology breakthroughs and apply that to how you can use LLMs to predict protein structure and function and encode protein data computationally. Attendees will learn techniques for how to use NVIDIA BioNeMo™, a generative AI platform for drug discovery, to simplify and accelerate training of models on their own data, ensuring easy and scalable deployment of models for drug discovery applications.

*Please choose either Session 2a or Session 2b.

Optimizing and deploying LLMs on self-managed hardware—whether in the cloud or on premises–can produce tangible efficiency, data governance, and cost improvements for organizations operating at scale. We'll discuss open, commercially licensed LLMs that run on commonly available hardware and show how to use optimizers to get both lower-latency and higher-throughput inference to reduce compute needs. Join us and learn how to scale up self-managed LLMs to accommodate unique business and application requirements.

*Please choose either Session 3a or Session 3b.

Cybersecurity is a data problem, and one of the most effective ways of contextualizing data is via natural language. With the advancement of LLMs and accelerated compute, we can represent security data in ways that expand our detection and data generation techniques. In this session, we’ll discuss advancements in LLMs, including how to leverage them throughout the cybersecurity stack, from copilots to synthetic data generation.​

*Please choose either Session 3a or Session 3b.

In this session, we'll answer any additional questions that attendees may have, beyond those discussed during the sessions.