Proof of concept of an adaptive learning system for engineering students
| Reference number | |
| Coordinator | Crash Course Sverige AB |
| Funding from Vinnova | SEK 1 694 180 |
| Project duration | May 2024 - November 2025 |
| Status | Ongoing |
| Venture | Banbrytande tekniklösningar |
| Call | Groundbreaking and scalable technology solutions in 2024 |
Important results from the project
The goals were partially achieved. KAI was broadened, from ALS to course assistant, reaching more students. Technical infrastructure delivered: ETL pipeline, a knowledge graph and AI-recommendations. Parts in production (30,000 AI chats, 14,000 AI-augmented solutions), others in validation. The partnership with Sveriges Ingenjörer enables deployment across all engineering programs, hence more efforts on a scalable infrastructure. A bottom-up approach for the recommendation system was favored.
Expected long term effects
The scalable infrastructure for AI generation and ETL pipeline means that KAI can soon reach all engineering students in Sweden and contribute to better throughput than today´s 54%. A collective and reviewable AI-usage can reduce environmental impact by reusing curated material instead of redundant individual generations, and can counteract the spread of unvalidated AI content. The infrastructure enables expansion into other STEM fields and lifelong learning for professionals.
Approach and implementation
The project started with scoping and design of KAI. Tech stack was modernized for AI integration. Early KAI releases (chat, augmented solutions) provided real-world insights. Knowledge graph built as POC. ETL pipeline required more time due to heterogeneous exam formats and quality requirements. The partnership with Sveriges Ingenjörer catalyzed national scaling. Gradual deployment via studiepass strategy. Work carried out internally with VNTRS consultants and student feedback.