Connect With the Experts
Get answers to your most pressing questions from NVIDIA technology experts.
What Are Connect With the Experts Sessions?
These interactive sessions give you a unique opportunity to meet with the minds behind NVIDIA’s products and research to get your questions answered, in either a group or one-on-one setting. You can meet with experts on a first-come, first-served basis at any time during the session window.
What Sessions Will Be at GTC Washington, D.C.?
- See what it takes to stand up a long-running agent and how NVIDIA NemoClaw™ shortcuts it. Assembling a harness, a model, a secure runtime, policies, and credential handling from scratch means weeks of integration decisions before your agent does useful work. NemoClaw helps lay the foundation for packaging a harness, model, and secure runtime, so teams across industries go from zero to a governed long-running agent fast.
- Discover why long-running, self-evolving agents need a different security model. Agents that install skills, spawn subagents, and run for hours can't be secured from inside their own process. NVIDIA Open Agent Safety Platform with OpenShell™ is enforced from the outside with kernel-level sandboxing, credential brokering so agents never hold live secrets, and policies formally verified before the agent runs. Even a successful prompt injection has a contained blast radius.
- Dive into the technologies for building the knowledge layer for enterprise agents that reason, retrieve, predict, and act across structured and unstructured business data. The NVIDIA Agent Knowledge Foundation provides tools to ground, retrieve, and manage context for all types of workflows, maximizing the performance of your agents.
- Learn how the NVIDIA Agent Toolkit applies across industry verticals. NemoClaw, OpenShell, Nemotron™ Retriever models, and the suite of NeMo™ libraries come together under the NVIDIA Agent Toolkit umbrella for building long-running agents. The toolkit applies to use cases across healthcare, higher education research, AI factory management, energy simulation, and collaborative teams, among others.
- Understand how to optimize your agents in production, using a combination of libraries including NeMo Relay as a runtime, NeMo Compass for failure analysis, and NeMo Eval Author for extracting actionable insights to evaluate and improve on. Combining those with NeMo Gym to evaluate your agent provides an environment that enables the agent optimization loop.
- Explore best practices in defending LLM agents from attacks.
- Learn how NVIDIA OpenShell™ governs agent execution as a secure runtime, controlling which resources agents can reach, what actions they can execute, and where inference is routed.
- Find out how the NVIDIA Open Agent Safety Platform extends OpenShell™ to full-stack governance, with NVIDIA Sentry on NVIDIA BlueField™-4 delivering continuous monitoring and hardware-based enforcement.
- Go deeper with our Networking DPU team for the latest and greatest in NVIDIA solutions that allow unprecedented scale-out and data integration.
- Understand how NVIDIA Confidential Computing protects AI workloads and data in use.
- Learn how to bring real-world data into simulation and robot learning through real-to-sim workflows and models that convert videos into robot trajectories.
- Explore how agents help set up, run, and evaluate robot simulations in NVIDIA Isaac Sim.
- Learn how agents support scene and asset creation, training setup, and hyperparameter tuning in NVIDIA Isaac Lab.
- Understand how to deploy trained robot policies as graphs, incorporate safety checks, and validate behavior from simulation to deployment on NVIDIA Jetson.
- Understand when reinforcement learning is the right post-training approach for specializing open models for domain-specific tasks and agentic workflows.
- Learn how to design training environments, curate rollout data, and define reward signals that improve reasoning, tool use, and task completion.
- Explore how to evaluate post-trained models and balance task accuracy, token efficiency, and training cost using repeatable RL recipes.
- Explore how conversational AI and reasoning agents can improve patient engagement, clinical documentation, healthcare operations, remote monitoring, and clinical trial workflows.
- Discuss approaches to integrating and combining language, speech, and clinical context into agentic digital health systems.
- Learn how NVIDIA Nemotron, Nemotron Speech, NemoClaw, OpenShell, Guardrails can be used in these workflows.
- Bring your architecture, integration, and scalability questions to NVIDIA experts through interactive breakout discussions and open Q&A.
- Discover how a modular toolbox of GPU-accelerated geospatial operations makes agentic processing approachable for developers and agents.
- Learn how fast I/O and processing on large rasters and vector data make rapid experimentation possible where it wasn’t before.
- Explore agent workflows that include on-demand product creation, accelerated vector search, and deployment to NVIDIA Jetson™ at the edge.
- Understand how to choose the right open model based on architecture, modality, task fit, and deployment requirements.
- Learn how open weights, datasets, and training recipes support reproducible model development and post-training for your own domain.
- Explore how to combine model selection, data preparation, post-training, and evaluation into a practical workflow for deploying specialized models and agents.
- Get expert guidance on accelerating engineering solvers and simulation workflows with NVIDIA computing platforms and CUDA-X libraries.
- Discuss how to build and integrate physics AI models with PhysicsNeMo, including simulation data requirements, model validation, and deployment.
- Explore how NeMo and tool enabled agents can automate simulation setup, parameter studies, and results analysis, with appropriate validation and human oversight.
Join us on November 30–December 3 and be part of the NVIDIA GTC Washington, D.C. experience.