Advik Bahadur – an electronic and computer engineering student at TCD – was part of the four-person team behind Aegis AI, the winning project at the HackEurope event held in Stockholm on February 21.

The 30-hour hackathon brought together more than 1,000 young developers, designers and problem solvers across three cities, Dublin, Paris and Stockholm, with winning teams at each location receiving €5,000 in prize money.

Access cloud-based lLLMs without exposing confidential information

Their innovation allows users to access cloud-based large language models (LLMs), like ChatGPT or Gemini, without exposing confidential information.

As the use of AI tools accelerates across industries, sensitive information, including names, addresses, credit card information, and medical records, is increasingly being entered into cloud-based systems. This data, once shared, may be retained or used to train future LLM improvements, embedding the confidential information outside of the user’s control.

Aegis AI addresses this challenge by introducing a privacy-first gateway that scans information locally, automatically detecting and masking sensitive data before it reaches the external models. This allows users to benefit from powerful AI systems while minimising the risk of data exposure. 

The team won the prize for the best project at the event in Sweden for their privacy-protecting AI tool.

"It’s been an absolutely exhilarating few days in Stockholm," TCD's Advik Bahadur posted on his LinkedIn account.

'True test of your core abilities'

"HackEurope isn't just a hackathon. It is a true test of your core abilities. It challenges you to go beyond just writing code; pushing you as a thinker, an inventor, and a communicator, trying to solve complex problems under intense pressure.

"That’s why I am incredibly proud to share that our team won HackEurope 2026 Stockholm!

"Over 36 hours, we developed an end-to-end AI security tool designed to tackle the critical challenge of data privacy and PII leakage.

"We engineered an on-device data anonymisation pipeline that strips away sensitive information at the source. By running Google DeepMind’s Gemma 3 model locally via Ollama, our system intercepts and sanitises data before it ever leaves the user's environment, and then uses Gemma to reconstruct the output from the sanitised information.

"This allows users to safely leverage exceptional third-party cloud models, from sponsors like Anthropic and Google DeepMind, without compromising their data security, and without sacrificing speed or quality." 

The team, which met for the first time when they landed in Sweden for the competition, included Advik Bahadur (Trinity College Dublin, Electronic and Computer Engineering); Ege Ozguven, (a first-year Chemical Engineering student at the  University of Bath); Lukas Noel (KTH Royal Institute of Technology, MSc Information Technology); and Jing Liu (Uppsala University, MSc Industrial Analytics). 

Reflecting on the team’s success, Ozguven said: “Winning Europe’s largest student hackathon in Stockholm was an incredible experience. I arrived in Sweden without knowing anyone on the team, and 30 hours later, we had built Aegis AI.

“I love building things and figuring them out fast. Chemical engineering is my academic path, but most of my energy goes into startups and turning messy ideas into products that work.

'This feels like the starting point, not the finish line'

“What excites me most isn’t just the win, but where Aegis can go next. This feels like the starting point, not the finish line.” 

Following their win, the team are working with Acceler8, the university’s student entrepreneurial society, to refine the tool and pilot a beta version. They hope to test Aegis in real-world scenarios where staff regularly use AI tools and must handle confidential information responsibly.

Siobain Hone, graduate enterprise manager at the University of Bath, said: “Hackathons and student enterprise competitions can be a fantastic way of introducing entrepreneurship to students. The University of Bath has a great track record with student and graduate entrepreneurs, supporting them through student groups and through our Entrepreneurship Programme, Dragons' Den and Innovation Awards, which go to graduates to support them working on their business for a whole year post-graduation.”

Beyond the hackathon, the Aegis AI team are focussing on developing their technology into a scalable, privacy-first infrastructure layer ready for AI adoption. They also hope to build a community of organisations, founders and everyday users committed to embedding privacy safeguards into everyday AI use.