The AI stack is shifting fast — frontier models, agentic architectures, and the infrastructure underneath are all evolving in real time. But what's actually working in production vs. what's still hype?

Novita AI, together with FounderCoho, is hosting a high-energy after-hours event during NVIDIA GTC 2026, featuring speakers from NVIDIA, Google DeepMind, Alibaba Cloud Intelligence (Qwen/Wan), Hugging Face, Kilo Code, Kimi (Moonshot AI), MiniMax, OpenRouter, RadixArk, and Z.ai.

Two practitioner-level panels, plus a relaxed networking mixer with the builders who are actually shipping.

If you're a builder, researcher, founder, engineer, or operator building with AI — this is the GTC side event you don't want to miss.

What You'll Walk Away With:

🔧 Panel 1 — The Agent Stack: From Frontier Models to Production Systems

How do you go from a powerful model to a reliable, deployed agent? Hear from the people building the picks and shovels — covering model selection trade-offs, agent architectures that survive real-world load, infrastructure decisions that save you months, and what 2026 will bring that changes how you build.

📈 Panel 2 — From Models to Market: Distribution & the Agent Value Chain

Building is only half the battle. Learn how leading operators think about distribution moats, model-to-market strategy, and what it actually takes to deliver production-grade performance at scale. No theory — just hard-won lessons from teams shipping to real users.

Networking Opportunities

Connect with builders, researchers, founders, and AI practitioners in a vibrant, relaxed environment during GTC week.

🍹 Complimentary food and drinks provided!

🎟️ The event is free to attend, and no GTC ticket is required!

(RSVP may be approval-based due to venue capacity.)

Agenda

🎤 Featured Panelists

Ari Messer, Head of Partnerships @ Kilo Code
Ari leads partnerships at Kilo Code, where he helps users take advantage of the latest LLMs and inference technology to power every aspect of agentic engineering. Kilo is the top open-source coding agent, with over 1.5M Kilo Coders around the world.

Banghua Zhu, Co-Founder & CTO @ RadixArk
Banghua is the CTO & co-founder of RadixArk and works on the SGLang and Miles projects. He earned his PhD from UC Berkeley, advised by Michael I. Jordan and Jiantao Jiao, and is joining the University of Washington faculty—bringing a rare “research + systems + open infra” combination.

Jiantao Jiao, Director of AI & Distinguished Scientist @ NVIDIA
Jiantao is Director of Research & Distinguished Scientist at NVIDIA and a professor at UC Berkeley (EECS + Statistics). He previously co-founded and served as CEO of Nexusflow.ai (agentic LLM startup) through its acquisition by NVIDIA, and his work spans the full LLM/agent pipeline—from data and training to post-training and systems.

Jeff Boudier, Head of Product @ Hugging Face
Jeff leads product at Hugging Face, one of the world’s most influential developer platforms across models, datasets, and tooling. He previously co-founded Stupeflix (acquired by GoPro) and held product, growth, and business leadership roles—bringing deep experience building platforms that developers and enterprises actually adopt.

Jeff Fu, Head of Model Solutions, Qwen Model Lab @ Alibaba Cloud Intelligence
Jeff is Head of Model Solutions, Qwen Model Lab at Alibaba Cloud Intelligence. Jeff leads the commercialization efforts for the Tongyi model family, including the Qwen and Wan multimodal models.

Linda Sheng, Head of Global Business @ MiniMax
Linda is currently the GM of Global Business at MiniMax. Prior to MiniMax, she was a technology investor at Warburg Pincus, a major US Private Equity fund. Before Warburg, Linda was an early member of Softbank Vision Fund in Silicon Valley and Asia, focusing on AI and technology investing. She began her career as a consultant with McKinsey, specializing in semiconductor, software and Internet verticals.

Ravin Kumar, Senior Researcher, Gemini & Gemma Models @ Google DeepMind
Ravin is a Senior Researcher at Google DeepMind working on the Gemini and Gemma model series. He has 10+ years building applied generative and probabilistic models—including bespoke Bayesian systems for SpaceX—and is a core contributor in the PyMC ecosystem and co-author of Bayesian Modeling and Computation in Python.

Ron Shepherd, VP, Business Development @ OpenRouter
Ron leads business development at OpenRouter, working at the center of model distribution, marketplace, and ecosystem partnerships. He’s also been active as an investor/advisor across AI and games, with prior experience including Polygon, Unity, and Electronic Arts.

Sarah Wang, Global Partnerships @ Kimi (Moonshot AI)

Sarah works with enterprises, developers, and research institutions to expand the global ecosystem around Kimi’s large language models. Her work focuses on strategic collaborations, developer adoption, and bringing advanced AI capabilities to glo bal partners.

Yu Jin (Lou), Head of Dev Ecosystem @ Z.ai
Yu Jin leads developer ecosystem at Z.ai, owning the full loop from foundation models to product, go-to-market, and brand. Sitting at the intersection of model infrastructure, developer experience, and product strategy, she spearheads Z.ai's open-source initiatives, helping developers worldwide leverage state-of-the-art AI to power real-world agentic workflows.

Moderators:

Junyu Huang, COO @ Novita AI

Junyu Huang is COO at Novita AI, where he works across product, go-to-market, partnerships, and operations to help developers deploy and scale model inference and agent workloads. Previously, he co-founded Toma.com (backed by Y Combinator and a16z) and worked on the AI infrastructure team at Scale AI.

Jing Conan Wang, Founder, Deepvista | Co-founder, FounderCoHo | Ex-Google Deepmind
Jing Conan Wang is the founder/CEO of DeepVista, a "Founder OS" for executable playbooks. A former Google DeepMind Senior Researcher with a Ph.D. in Reinforcement Learning (BU), he also co-founded the FounderCoHo community.

📍 Reserve your spot now—space is limited!

Community Partners

Organizer

Novita AI: Novita AI is an AI & agent cloud platform helping builders serve and scale models and agentic workloads with strong performance, reliability, and cost efficiency.

Co-Hosts

FounderCoHo: FounderCoHo is a community where founders support each other by sharing knowledge and building connections. Starting a company is a brave endeavor. Many of us have faced numerous challenges and learned valuable lessons along the way. FounderCoho aims to be that supportive community where founders can share their stories and get fueled for their company-building journey.

Plug and Play Tech Center: Plug and Play is the largest open innovation platform that brings together the best startups, the world's largest corporations and seasoned investors. It aims to drive the future forward with corporations, governments, startups, investors, and universities.

Disclaimer

NVIDIA’s involvement in this event is limited solely to the technical insights provided by its invited speaker. This event is not sponsored or endorsed by NVIDIA. Any opinions expressed by NVIDIA representatives are their own and do not necessarily reflect official NVIDIA policy.

GTC After Hours: From Models to Agents to Infra