GLM 5 is now live on Fireworks. Try It Today.

Booth #631

Fireworks at Nvidia GTC 2026

Open-source AI models at blazing speed, optimized for your
use case, scaled globally with the Fireworks Inference Cloud.

San Jose, CA

March 16-19, 2026

Meeting Space #6053

Meet us at GTC

Meet the Fireworks AI team at booth #631

Stop by for a demo, collect limited-edition branded merch, and hear how Fireworks can provide the platform to build your Generative AI capabilities - optimized and at scale.

Check us out on stage

Building Self-Improving AI Agents: Use Production Data to Beat Frontier Models

Date: March 16
Time: 4:20 PM - 4:35 PM PT
Speaker: Roberto Barroso, Applied AI, Fireworks AI

This session presents a practical framework for building self-improving agents that are better, faster, and cheaper than frontier alternatives. We'll explore how to systematically integrate production data into your AI pipeline—whether through automatic prompt optimization, supervised fine-tuning, or reinforcement learning—to build specialized agents that excel at your specific tasks.

Check out demos at our booth

Meet with product experts and stop by for our 10-minute demos

Fed-AI-Savant: Federal Reserve Knowledge Assistant with Function Calling:

  • An AI-powered knowledge assistant that answers questions about Federal Reserve minutes and economic outlook. The demo showcases function calling capabilities using Fireworks AI's optimized inference platform, demonstrating how LLMs can reliably call APIs to retrieve and synthesize information from complex policy documents. Users can ask natural language questions about monetary policy and receive accurate, contextual responses.

Eval Protocol: Reinforcement Fine-Tuning for Production Agents:

  • A live demonstration of reinforcement learning fine-tuning (RFT) using Eval Protocol, an open framework for training agents across any language, container, or framework. The demo showcases training an SVG generation agent end-to-end: the model generates SVG code from text prompts, visual evaluation via GPT-4 scores fulfillment of requirements, and Fireworks RFT improves the model iteratively. Attendees see dramatic before/after comparisons as the model improves through training epochs.

Book a meeting with the Fireworks AI team

Run and tune your open-source AI models on our highly scalable, optimized virtual cloud infrastructure. We have product experts, engineers, and executives on-site and available for meetings and demos.

Meetings will be hosted in our meeting space #6053 and booth #631.

Join our Fireworks Afterparty on March 17

Fireworks: GTC Afterparty

Join Fireworks AI for an exclusive evening bringing together AI leaders and builders shaping the future of generative AI.

We’re hosting a curated group of founders, engineering leaders, and AI practitioners for drinks, bites, and thoughtful conversation. A quick walk from San Jose Convention Center, join people building real systems, deploying real models, and pushing the boundaries of what’s possible with AI infrastructure.

Space is limited and RSVP is required.

Start building with $100 of Credits

How to Redeem Free Credits

To make it easy to start building on Fireworks today, we're giving all attendees of Nvidia GTC 2026 $100 worth of free credits. To claim your credits, simply create an account and enter the code GTC2026 from within your account. For full instructions, click the "Claim $100 in Credits" button to watch a ~1 min explainer video.

Get stuck or have questions? Stop by our booth at #631 and we'll be happy to hep out?