Build agents that do real work
It’s easy to build an AI agent that looks good in a demo. It’s incredibly hard to build one that work in real environment.
Across three distinct tracks—Building 24x7 autonomous agents, Building for Agents, and a General track—you will build the underlying architecture to cross the last mile of agentic AI. Join us to build agents that don't just execute tasks, but execute them exactly as the real users the actions belong to.
Date: Saturday, June 6thTime: 09:30 AM – 7:30 PM
The problem
Most AI agents today are trapped in the "demo zone." They work perfectly on a carefully curated golden path but break the moment they encounter a real-world edge case, an unpredicted API response, or ambiguous context. They lack long-term memory, fail to reason through complex workflows, and struggle to authenticate or act safely on behalf of real users. The gap between a viral social media demo and a production-grade agent that can reliably sit in on meetings, manage workflows, and operate autonomously without constant supervision is massive. This hackathon is about bridging that gap.
Tracks
Building 24x7 autonomous agentsBuilding for AgentsGeneral track
Schedule
🕘 09:30 – 10:30 AM Check-InGrab your badge and get settled in for the day.
🕤 10:30 – 11:30 AM Welcome & Track OverviewsKickoff, API walkthroughs, and track breakdowns.
🕙 11:30 AM – 4:30 PM Build TimeHack, attend workshops, get mentor guidance. 🍕 Pizza lunch provided.
🕓 4:45 PM Submissions DueFinal code locked. No extensions.
🕔 5:00 – 5:25 PM Community showcases4 companies, 5 mins each — tools and products the room should know about
🕔 5:30 – 7:00 PM — Demos & JudgingPresent to the room and vote for your favorites.
🕖 7:00 – 7:30 PM — Networking & DrinksCelebrate, connect, and decompress.
Prizes
Credits and gadgets worth $50K shared between 3 tracks.
Judging criteria
Projects are scored equally across four dimensions:
🔭 Innovation (25%) — Is the approach novel? Does it solve the problem in a way we haven't seen before?
⚙️ Technical Complexity (25%) — How well-built is the solution?
🌍 Impact (25%) — Does this solve a real problem? Could it scale beyond the demo?
🎯 Presentation & Usability (25%) — Is the demo clear and compelling? Would someone actually want to use this?
Meet the judges
Manoj Bajaj - Founder of Authsome, building open-source trust infrastructure for autonomous agents. An IIT Delhi graduate with experience across HFTs, hedge funds, and AI infra, he has spent years building high-performance systems before moving into agentic software.Vedanth - Vedanth is a software engineer with experience across Uber, Microsoft, and Slice, where he built and scaled systems handling payments and user notifications at scale. He is currently co-founding BrowserWire, a browser automation infrastructure startup building developer-facing tools for human-like, anti-bot-evasion web sessions.Abhimanyu Dhamija - Abhimanyu is serial entrepreneur and Tech executive. He has built Amber and KoalaLab as founder and CEO and earlier as an executive built Fintech Product at Khatabook and ran data sciences at Housing.com. He is also the creator or open-source tools BOLT and PINNY and graduated from IIT-Bombay with degrees is computer science and statistics.Shubham - An IIT Kanpur and UChicago graduate with experience spanning Goldman Sachs and scaling a three-person systematic hedge fund to a multi-billion dollar book within 18 months. With expertise in AI/ML infrastructure, he is currently working on an AI startup in stealth.
Who should come
This is for you if you're a founder, hacker, AI engineer, or builder who wants to work on the leading edge of agentic AI. No experience with meeting APIs required — we'll have API docs, live walkthroughs, and mentors on hand all day.
Whether you come solo or with a team, you'll leave with working code, new connections, and a sharper intuition for building agents that actually do things.