2 months ago
Meet more judges
Hi Everyone!
Today, I'm introducing you to a trio of guest judges that I'm excited to have join us. They are knee-deep in AI, and I love the perspectives they've brought in. Check out their companies, they are each doing some pretty cool things.
Meet Deepanjan Mukherjee, Director of Engineering, athenahealth
Deepanjan leads a 40-person team building patient experience platforms serving 10M+ users monthly. Published researcher on multi-agent security frameworks. Carnegie Mellon alum who led athenahealth's Telehealth launch during COVID-19.
What excites you most about what AI agents are becoming?
We're moving past AI as a tool you prompt and toward AI as a collaborator that acts. What excites me most is agents that can handle the unglamorous work — authentication flows, API negotiation, and context management across services — so builders can focus entirely on the problem they're actually trying to solve. Token Vault is a great example of infrastructure that makes that real.
What kind of submissions are you hoping to see? What would make a project stand out to you?
I want to see agents that do something genuinely useful and handle failure gracefully. The projects that stand out won't just demo well — they'll show the builder thought through edge cases, security boundaries, and what happens when the agent hits an unexpected state. Production-thinking at hackathon speed is rare, and I'll be looking for it.
What advice would you give to builders who want to push the boundaries of what agents can do?
Pick one hard problem and go deep on it. The most impressive hackathon submissions I've seen don't try to do everything — they find the one thing their agent does better than any human workflow and make that undeniable. Let Token Vault handle the auth plumbing so you can put all your energy there.
---
Meet Nihal Kaul, Lead Software Engineer, Revscale AI
Nihal Kaul is a Lead Software Engineer at Revscale AI who builds scalable, cloud-native systems. He works across distributed systems, infrastructure, and reliability, with a focus on applied AI development, building persistent, context-aware agents that make products more adaptive and useful.
What excites you most about what AI agents are becoming?
What excites me most is that we’re moving from AI as a passive tool to AI as an active collaborator. Agents are beginning to plan, reason, and interact with software systems in ways that resemble real teammates, which opens up entirely new ways to build products and workflows.
What kind of submissions are you hoping to see? What would make a project stand out to you?
Projects that go beyond simple prompt wrappers and demonstrate thoughtful system design stand out the most. I’m excited to see teams build agents that interact with real systems, coordinate tasks, and solve meaningful problems with clear user value.
What advice would you give to builders who want to push the boundaries of what agents can do?
Focus less on the model and more on the system around it. The most interesting agent systems combine memory, tools, and structured workflows to create something reliable and useful. The teams that think deeply about how agents interact with real-world systems will build the most compelling projects.
---
Meet Paul Chuang Li - Co-founder & CEO, OpenBuilder
Co-founder of EasyCode, a coding extension used by 1.5M+ developers across JetBrains and VS Code. Deep in the agentic AI space, building tools where agent-to-API auth patterns are a daily reality.
What excites you most about what AI agents are becoming?
AI agents are moving from passive tools to systems that can actually take action. What’s exciting is when they can operate across tools, run workflows, and continuously make progress without constant human input. That’s when they stop being demos and start becoming genuinely useful.
What kind of submissions are you hoping to see? What would make a project stand out to you?
I’m most interested in projects where agents are doing things they can do better than humans: not just augmenting humans, but actually taking ownership of parts of a workflow.
What stands out is when an agent meaningfully reduces effort or unlocks something that wasn’t practical before.
What advice would you give to builders who want to push the boundaries of what agents can do?
Focus less on making the agent sound smart, and more on making it useful. Start from a real problem, then design the agent around taking actions and delivering outcomes. The teams that stand out usually keep things simple, integrate where it matters, and actually get something working end-to-end.
One more thing...
Using VS Code and want to manage Auth0 with AI directly in the IDE with the Auth0 MCP Server? Here's a nice explainer video on youtube.
Questions?
If you have any questions about the hackathon, please post on the discussion forum.