DeepLearning.AI

At AI Dev 25 x NYC, Robert Crowe, Product Manager at Google showed how Flax NNX makes JAX far more intuitive for building and training neural networks. He walked through how JAX can automatically distribute models across hardware, making it easier for developers who are just getting started. Robert also highlighted why efficiency matters: accelerators are costly, and tools like roofline analysis help teams get the most out of them.

At AI Dev 25 x NYC, Hatice Ozen (Head of Developer Relations, Groq) showed how compound AI systems can build deep-research agents with a single API call. She walked through how agents choose tools, reason over results, and loop until they reach an answer — and why latency becomes the real bottleneck. Groq’s LPU architecture is designed to keep this workflow fast enough for real applications.

Stefano Pasquali, Domyn's Head of Financial Services, at AI Dev 25 x NYC: "Domyn believes in building transparent, auditable and sovereign AI ecosystems, where innovation meets accountability."

During the panel "Breaking the Limits of AI Growth," Kay Zhu, CTO at Mainfunc (Genspark), explained how an AI-native approach helps avoid development bottlenecks. By focusing on what AI is good at, and adapting products accordingly, they’re able to launch new features every week. Follow along for more highlights from AI Dev 25 x NYC!

Andrew Ng kicked off AI Dev 25 x NYC by explaining why AI continues to accelerate: coding is getting faster, teams can prototype far more quickly, and the real bottleneck is now gathering user feedback. He closed by encouraging attendees to connect, collaborate, and build together, just like the conversations that helped spark AI Aspire, after meeting Kirsty Tan at the previous edition of AI Dev 25.

SAP's Christoph Meyer and Lars Heling described why agents fail inside complex enterprise systems. “AI agents might struggle in complex systems for two reasons — choosing the correct API to execute, and understanding the business process context.” Lars emphasized that “APIs do not exist in a vacuum, they happen in a discrete order at different times.” They explained how knowledge graphs solve this by defining semantics through ontologies: “Resources, APIs, and business processes all become different nodes in the knowledge graph.” More session insights up next.

Turing is at AI Dev 25 x NYC highlighting how they help frontier labs and companies move from research to deployment with high-quality data operations, talent, and tooling. Visit them at booth 22 to learn how they support AI at scale!

At the panel “Software Development in the Age of AI,” Malte Ubl, CTO of Vercel described how AI is changing team workflows. PMs and developers can now align around working prototypes from day one, reducing misalignment and speeding up iteration. He also noted how agents can investigate real issues from tickets and gather the context developers need—reshaping what “human-in-the-loop” means. Malte added that focused vertical teams are better positioned to build successful products than broad AI labs spread across many directions. More AI Dev 25 takeaways coming up!

Andrew Ng kicked off AI Dev 25 x NYC! "At the last conference in San Francisco, I met and talked to Kirsty Tan, who ended up founding AI Aspire. Here at the conference, I hope you'll all meet each other, make friends, and make lasting connections." Stay tuned for more AI Dev 25 updates throughout the day!