Case study TriSE – Trends in Software Engineering
In the TriSE project, fortiss developed a scalable, semi-automated platform for Siemens to systematically analyse trends in software engineering, as the evaluations previously carried out manually were too slow and too limited. The new solution continuously integrates data from open sources such as scientific publications and developer platforms and uses machine learning methods to evaluate this information in a structured manner and convert it into strategically useful insights.
The result is a powerful tool that makes technological developments transparent, identifies emerging trends at an early stage, and supports both idea generation and research and innovation planning at Siemens. This strengthens the company’s ability to identify relevant software trends more quickly, evaluate them systematically, and incorporate them in a targeted manner into future development and decision-making processes.
Challenge
Until now, Siemens had primarily identified new trends in software engineering through manually conducted interviews. This approach was neither scalable nor fast enough to respond to new developments, and allowed for only limited involvement of experts.
Furthermore, the existing process was characterised by outdated data, limited coverage of relevant subject areas and long feedback cycles.
Solution
fortiss developed an automated, data-driven system (TriSE) that continuously updates trend analyses whilst enabling broader and more efficient integration of specialist knowledge.
To this end, the TriSE platform integrates public data sources such as arXiv, GitHub and StackOverflow and, using machine learning methods, provides trend analyses, knowledge graphs and intelligent search functions. The platform comprises three central modules – Relations, Statistics and Q&A – and is based on technologies such as Word2Vec, BERT, Neo4j and Elasticsearch.
In addition, a user-friendly interface was created, enabling Siemens employees to define and explore topics and develop them further in a targeted manner using generative AI. Through iterative development steps closely coordinated with Siemens, the platform was continuously expanded and enhanced with real-time data processing, AI-supported knowledge enrichment and practical analysis functions.
Result
- Development of a fully automated Extract, Transform, Load (ETL) pipeline using Apache Airflow to capture, cleanse and enrich various software engineering data sources
- Introduction of regular trend updates and quarterly update cycles to improve the relevance and accuracy of results
- Development of a semi-automated knowledge module based on large language models to enable faster expansion of subject areas
- Provision of an intuitive user interface for defining and managing subject areas, including the option for expert feedback
- Integration of analysis and visualisation tools to identify high-growth technologies, categories and their interrelationships
- Introduction of a ‘Promotion View’ that allows irrelevant terms to be hidden and relevant technologies or topics to be specifically highlighted
- Expansion of the platform to include analyses of technology maintenance and activity trends (e.g. GitHub data, OpenSSF scores, dependency graphs)
- Provision of practical trend information as a basis for future AI-supported idea generation and innovation planning at Siemens
Outcome
The project demonstrates that a scalable, AI-powered trend analysis platform significantly speeds up the identification of technological developments whilst also making it more robust. This makes innovation processes at Siemens more efficient and ensures that strategic decisions are based on a broader and more up-to-date knowledge base.
Project partner
