We Took Zerve to Hackathons Around the World. Here's What Happened

Over the past few months, we've been showing up at hackathons around the world and handing people Zerve to see what they'd do with it. We went to university campuses in India, longevity research sprints in Berlin and Paris, and online competitions with thousands of participants.

What came back surprised us. Not because the projects were impressive (they were) but because of what participants said about the process. 

Zerve x Unstop Hackathon

We partnered with Unstop for a hackathon that drew competitors from across India, and the winning projects went well past the experiment stage into something you could actually deploy.

Ruturaj Sonkamble won first place. He built an autonomous bidding engine that turns ad spend into profit, showing a 12.8X ROI improvement and dropping cost per acquisition from $15.70 to $1.72. "I cannot stress enough how useful Zerve was for this," he said. "It bridges the gap between 'Data Science Experiment' and 'Production Microservice.'"

Second place went to Poornasri P, a software engineer at OneBill. She built and deployed a subscription revenue forecasting system for financial planning, hiring decisions, and marketing allocation. Her words: "This project goes beyond notebooks. It's a real analytical system built and deployed on Zerve."

Durgesh Bhardwaj, a software engineer at Innefu Labs, took third with a customer churn prediction system. The part that stood out: he designed it so non-technical users could actually interact with it. That last detail matters more than people think.

Zerve AI Datathon at Techfest, IIT Bombay

Techfest is Asia's largest science and technology festival, and we ran our AI Datathon inside their competition lineup. Participants had to predict health insurance claims from a messy, highly imbalanced dataset. Scoring was based on Normalized Gini, which rewards models that rank claims correctly rather than just predicting averages.

350 students entered the qualifying round. 90 made it through, spread across 29 teams. They had the Zerve agent at their disposal, and we saw teams use it to handle preprocessing and validation steps that would have eaten hours otherwise.

First place went to Team Pancakes: Ishansh Sharma, Ridhwan Deshwal, Shivaay Dhondiyal, and Akansh Gupta.

Fusion Forge (Neeraj Parekh and Aairah Zutshi) came in second.

Two teams tied for third: Team K (Bhoomi Kaushik, Manya Gupta, Mrityunjay Srivastava, and Shreeyam Bangera) and Team Bharat Bytes (Vedang Mendhurwar, Ruturaj Rajwade, and Harshal Pednekar).

Thanks to our judges Kreshnaa Raam (Lead Data Scientist at Zerve) and Ganesh Shyam Arichandran (Data Scientist at Zere) for putting in the hours with a tough dataset.

Longevity Hacks: Berlin and Paris

Longevity Hacks runs global innovation sprints where AI builders tackle aging research. We sponsored two of their events.

Berlin came first in November 2025 at Delta Campus, where 120 people showed up for a 36-hour build session. The brief was to create AI agents that design personalized longevity protocols and accelerate aging research. We ran a virtual workshop on data analysis agents beforehand. 

Paris followed a week later in November 2025 at Alpha Intelligence Capital. About 60 builders participated, roughly half the size of Berlin, though the projects were more technically complex. Teams focused on precision medicine, building systems that could tailor longevity protocols to a specific person's bloodwork, habits, and health history. 

What made both events interesting was the mix of people in the room. AI engineers sat next to clinicians. Data scientists paired up with biotech founders. That kind of cross-pollination is hard to manufacture, and Zerve's flexibility as a shared data science environment helped teams with wildly different skill sets find common ground.

Zerve.AI 2026 Hackathon on HackerEarth

This one is still running. We put together a $10,000 data challenge on HackerEarth that asks a deceptively open-ended question: what drives successful usage of a data platform?

We gave participants a sample of real Zerve application event data, everything from canvas creation to code runs to deployments. The core question is which user behaviors and workflows are most predictive of long-term success, and participants get to define what "success" means. Some have gone after depth of analysis, others reproducibility, others workflow complexity.

All development has to happen inside Zerve, and all results have to be reproducible within a Zerve project. You pick the methods. As long as your work lives in Zerve, the approach is up to you. 

First place wins $5,000. Second and third take home $3,000 and $2,000. Winners also get a project spotlight on our channels, Zerve swag, and a year of Pro access.

Submissions are open through March 30, 2026. Signing up is free and you get Zerve credits the moment you register at zerve2026.hackerearth.com.

What People Told Us

We heard the same thing at almost every event, phrased differently each time but pointing at the same idea.

Students who had never touched certain modeling techniques could describe what they wanted in plain language and watch the Zerve agent build it. One team at Techfest told us they tried three different ensemble approaches in the time it would normally take to set up one. At Unstop, Ruturaj's bidding engine went from concept to deployed system inside the competition window.

But the comment that stuck with us most came up in conversation after conversation: other tools let me experiment, but this one let me finish. When you have 36 hours or a weekend deadline, the distance between a working notebook and a deployed system is everything.

The range of problems people solved tells the story: ad bidding, insurance claims, subscription revenue, customer churn, personalized longevity protocols. These are totally different domains, but the common thread was that every winning project could be handed off to someone who doesn't write code, and it would still work.

Get Involved

Our HackerEarth hackathon is open through March 30, 2026, with $10,000 in prizes, real data, and a real leaderboard. Want to know where we'll be next? Check out our upcoming hackathons on Luma to stay in the loop.