Make AI visible: How Professor Pekka Abrahamsson brings research into shipping code, student work, and everyday decisions
Published: 22.12.2025 - Tampere University
Pekka Abrahamsson, Professor of Software Engineering at Tampere University and AI Researcher of the Year 2025, is pushing AI out of papers and into practice. His focus is clear: build useful systems with companies, create hands-on learning for students, and connect top-tier research with local needs.
"A guiding principle throughout my career has always been that science must also have an impact," he says. That mindset runs through his work from Oulu to Italy, Norway, and back to Finland-now anchored at the University Consortium of Pori.
From Pori to production: data in, decisions out
Abrahamsson sees Pori as a place where AI moves fast from idea to demo to deployment. The region has capable companies with real needs in data, software, and decision support. The approach is straightforward: bring company data into AI securely, explain insights clearly to both executives and frontline teams, then build demos that prove value.
He sees his role as a catalyst-connecting companies, researchers, and funding, and making collaboration easy to start.
GTP-Lab: students and doctoral researchers at the core
Founded in 2023, the GTP-Lab is Abrahamsson's research group focused on strengthening Finland's position in software development and AI. With teams in Pori, Tampere, and Seinäjoki-and partners in Sweden, Norway, Brazil, and Tanzania-the lab is scaling fast.
This isn't passive research. Students and doctoral researchers build real AI systems with companies-software, data pipelines, and decision-support experiments. AI helps with development and testing, including code in legacy languages few people use anymore.
There's also a clear warning: AI can hallucinate. So AI literacy-critical thinking, verification, and good prompts-has become a basic civic skill.
How he works: casual tone, serious goals
Abrahamsson doesn't fit the old-school professor stereotype. He's open about risk, exploration, and ideas that don't come pre-packaged with answers. "My strength is that I don't say no. I know the word, but I don't use it."
The culture he's building rewards action: test ideas quickly, form interdisciplinary teams, and build what wouldn't exist otherwise.
Projects worth a developer's attention
- ANSE (AI Native Software Engineering) - A joint initiative with the University of Jyväskylä to build AI-native software engineering methods. The aim: make AI an integral part of the development process, not just another tool. Funded through Business Finland's Näytönpaikka pilot programme. Business Finland
- EVIL-AI - A four-year project on the negative impacts of AI agents and how to prevent them. Led by Professor Henri Pirkkalainen with Abrahamsson and Assistant Professor Johanna Virkki. Focus: agents that can deceive, coordinate, and act outside intended bounds-and how to detect and mitigate those risks early. For practical risk frameworks, see NIST AI RMF.
- AI Champion - Business Finland-funded project on AI agents in construction and building services engineering. Teams span Built Environment, Management and Business, and ITC, with the University of Oulu and companies including Granlund, Fira, and Koja Chiller. Core question: why build centralized, monolithic systems if AI agents can stream and enrich data across existing silos? GTP-Lab is the tech backbone, with an emphasis on domestic data and AI sovereignty.
- AI Newsroom - A fully AI-driven 24/7 newsroom prototype that modeled roles like editor-in-chief, news editor, and journalist. Outcome: clear recommendations on where AI helps, where it fails, and where human judgment stays non-negotiable. A practical meeting point of software engineering and journalism.
Practical takeaways for engineers and tech leaders
- Move work toward AI-native flows: planning, coding, testing, and ops with AI in the loop from the start.
- Build secure data connectors and permission models first; without clean, governed data, your AI stays a demo.
- Prototype quickly with company data. Ship small demos that show decision impact for both executives and frontline teams.
- Treat agents as software you threat-model: logging, isolation, rate limits, and rollback plans.
- Use AI for legacy code: translation, refactors, tests, and documentation-then review like your uptime depends on it.
- Question monoliths. If agents can orchestrate across systems, you might get agility without rebuilding everything.
- Make AI literacy mandatory: validation habits, prompt hygiene, and clear escalation when outputs look suspicious.
Why this matters
This is research you can feel at work. Better software practices. Faster demos. Tighter loops between data and decisions. And students who graduate already shipping useful AI.
If you're building AI into products or processes and want structured ways to upskill teams, explore focused learning paths for developers and data roles: Complete AI Training - Courses by Job.
Your membership also unlocks:
