Mearra’s CCO, Ulla Koho, and Head of Technology, Tomi Mikola, recently shared the details of a project innovated and built by Mearra: an AI-powered search for the University of Helsinki. Presented at Drupal Developer Days in Athens, the session focused on how large institutions can solve the problem of fragmented information. In this article, we highlight the project’s core technical pillars and provide a link to the full session recording.
The Problem: Information Exists, but Access Fails
The University of Helsinki manages content for 35,000 students across seven different upstream systems, including intranets, helpdesks, and curricula. The challenge was not a lack of information, but the difficulty users faced in finding it.
As Ulla Koho explained:
Most organisations don’t have a content problem — they have a clarity problem. The information exists. But when people can’t find the right answer quickly, confusion becomes expensive: emails pile up, support teams get overloaded, and trust erodes quietly. The organisations that win aren’t the ones with the most content. They’re the ones where people get to the right answer at the right moment. Clarity isn’t a nice-to-have. It’s a competitive advantage.
The project’s goal was a strategic shift: moving from content management to answer delivery.
The Solution: Built by Mearra in 90 Days
Our team — consisting of three engineers and two service designers — built and launched a production-ready “answer layer” in just three months. This solution was designed to provide students with answers they can immediately act upon.
Key Technical Facts of the Mearra Solution
Answer Layer: A custom stack utilizing Node, Fastify, and LLM technology sitting on top of existing data.
Hybrid Retrieval: The system combines semantic vector search with traditional Elasticsearch. This ensures the search can handle both broad natural-language questions and specific keywords like course codes or teacher names.
Trust Through Grounded Generation (RAG): Using Retrieval-Augmented Generation, the AI only provides answers based on the University’s authoritative data. If the information is missing, the system is programmed to refuse to generate a fake summary.
Direct Source Visibility: To ensure reliability, every claim made by the AI includes a direct link to the source URL, allowing users to verify the information.
Automated Data Pipelines: We maintained the University’s existing CMS structure by automating 42 different data pipelines and cron jobs to keep the AI’s knowledge base updated.
The Result: Measurable Clarity
By implementing this search layer, Mearra made the University’s content quality measurable for the first time. Through an admin dashboard, content gaps are now visible, and ownership of specific information is traceable.
This project reflects our commitment to our slogan: Business. People. Tech. Together. We focus on high-impact technology that serves human needs and institutional goals — and build all that in close collaboration with our clients.
Watch the full presentation from DevDays Athens below to see the demo and the technical architecture behind the solution.
Watch the recording: AI Search as Digital Public Service
We gladly discuss the opportunities of AI search in your organization. Send Ulla a message or click the Contact us button below and we will get back to you soon.
