AERIS | AI-powered retrieval and synthesis of engineering knowledge

Region:

Brussels

Financed by

Turning fragmented engineering documentation into accessible knowledge

Engineering companies generate vast amounts of technical documentation, but valuable knowledge often remains scattered across reports, drawings, databases and legacy files. AERIS (Automated Engineering document Retrieval and Intelligent Synthesis) develops generative AI solutions that help engineers reliably retrieve, synthesise and reuse this information more efficiently while ensuring data confidentiality through secure on-premises deployment.

Target group

The project is relevant for organisations managing large volumes of technical documentation, including:

Engineering consultancies and owner-engineers
Industrial companies with extensive technical document repositories
Organisations operating under strict data confidentiality or on-premises deployment requirements
Engineering teams responsible for compliance, reporting and knowledge management
Companies looking to accelerate the onboarding of new engineers
AI and data teams working on domain-adapted document intelligence solutions

Context

Engineering organisations rely on decades of accumulated technical knowledge stored in reports, drawings, specifications, images and legacy databases. Finding the right information often requires engineers to manually search, compare and consolidate data from multiple sources, making these activities both time-consuming and error-prone.

The challenge is even greater in environments where sensitive data cannot be transferred to the cloud. Existing document intelligence solutions rarely combine advanced multimodal understanding with secure on-premises deployment and the ability to reason across large collections of engineering documents.

Objectives & results

AERIS develops AI-powered methods that enable engineering organisations to retrieve, interpret and synthesise technical knowledge more efficiently while maintaining full control over sensitive data.

The project aims to:

Improve the understanding of complex engineering questions, including ambiguous and multi-step queries
Combine information from technical documents, images, tables and other heterogeneous sources
Generate reliable, reference-based answers using advanced retrieval-augmented generation (RAG) techniques
Validate scalable AI solutions that operate efficiently within secure on-premises environments

Key expected results

AI methods capable of understanding and decomposing complex engineering queries
Advanced retrieval strategies combining metadata, graph-based reasoning and vector search
Techniques for extracting and representing multimodal engineering content
A validated prototype tested on real-world engineering documentation
Benchmark datasets and evaluation methodologies supporting future industrial AI developments

Approach

Within AERIS, funded by Innoviris, Sirris collaborates with Tractebel Engineering. Tractebel provides the industrial use case, technical documentation and validation environment. Sirris contributes its expertise in Generative AI, multimodal language models and Retrieval-Augmented Generation (RAG).

Together, the partners develop methods to efficiently and reliably retrieve, combine and synthesise information from complex engineering documentation. The resulting solutions are designed to be applicable across a wide range of industrial sectors.