UK university students build RoomQuest and MeetSpot at Google Gemini AI hackathon

The London event selected 52 participants from more than 900 applicants for a one-day application development and pitching challenge

Google’s UK Student AI Hackathon brought selected university students to London for a one-day prototype challenge. Photo credit: Harmehar Kaur

Google brought computer science and engineering students from UK universities together in London on July 24 for a one-day AI hackathon focused on building working applications with Google AI Studio and presenting them to a panel of Google judges.

Harmehar Kaur, a computer science student at King’s College London, was part of a five-person team selected for the event.

“Out of nearly 1,000 applicants, our team of five was selected to build an AI solution using the Gemini API and present our working prototype directly to a panel of Google experts,” Kaur wrote on LinkedIn.

University of Warwick computer science student Varshnay Sinha gave a more precise participant figure in her own LinkedIn post: “Just attended the Google UK Student AI Hackathon, selected as 1 of 52 builders from 900+ applicants, and built MeetSpot in 4 hours!”

Eligibility was restricted to undergraduate students in their second or third year of computer science or a related subject at a UK-based university. The in-person event ran from 9 a.m. to 6 p.m. at 6 Pancras Square in London.

Participants moved from an ideation session and mentor feedback into prototype development, submission, presentations and judging. Google engineers provided technical and product guidance, while projects were assessed on technical execution, product design, innovation, teamwork and realistic business impact.

RoomQuest turns physical spaces into AI-generated games

Kaur worked with Aarian Malhotra, Zidane Imran, Daniel Chigbu and Quvonchbek Rahmonov to develop RoomQuest. The team began with a question about whether a digital game could prompt more interaction with the player’s immediate surroundings.

“In today’s disconnected world, how do we modernise gaming and encourage people to look up from their screens to interact with the physical world?” Kaur wrote.

The team’s answer was RoomQuest, which uses images of a player’s surroundings to generate an escape-room game.

“To answer that, we built RoomQuest - an AI-powered game that turns any physical room into a personalised escape-room adventure,” Kaur explained.

Players upload photographs of their surroundings, allowing Gemini to identify physical objects and create a story with rhyming clues. Describing the process, Kaur wrote: “Gemini identifies real physical objects in the room and generates a custom story with rhyming clues.”

Players then photograph the objects they believe solve each clue. According to Kaur, “Gemini Vision verifies in real time using structured outputs.”

The team built the prototype with React, TypeScript, Node.js, Express, Vercel and the Gemini API. Thierno Thiam mentored the group.

“Designing, integrating, and deploying a functioning product within such a tight timeframe was an incredible lesson in fast-paced decision-making and collaboration,” Kaur wrote. “Beyond the project, hearing from Google employees and interns about their career journeys provided invaluable insight into the tech industry.”

MeetSpot weighs travel, access and weather

Sinha used the four-hour build period to develop MeetSpot, an application intended to help groups choose where to meet rather than simply provide directions to a predetermined destination.

She framed the gap in existing journey-planning tools directly: “Google Maps can tell you how to get somewhere. It can't settle where a group should meet.”

Users enter their starting points and the prototype searches for places intended to provide fairer journey times across the group. Sinha said the recommendations can also account for dietary and accessibility requirements, budgets, atmosphere and conditions on the day.

“MeetSpot does it in seconds,” Sinha wrote. “Everyone drops in their starting point and it finds real places with fair journey times for the whole group, with dietary and accessibility requirements, budgets and atmosphere already accounted for, and the conditions on the day factored in so nobody ends up on a terrace in the rain.”

The prototype can use an uploaded Google Maps Timeline export to personalize its suggestions. Sinha also made a specific privacy claim about that feature: “You can also upload your Google Maps Timeline export so it learns the sorts of places you actually go, processed in your browser and never uploaded.”

“Gemini is the reasoning engine, working over Places, Routes, Weather and Timeline data,” she added.