Mechanical Engineering
Data science and AI for the automotive domain – in development, production and quality assurance.
Supper & Supper builds data science and AI solutions for OEMs, suppliers and machinery manufacturers.
The data science core, embedded in your systems
We deliver the data science and AI core of an initiative: data preparation, models, evaluation logic and the integration into the existing landscape – in collaboration with the customer's service providers, tool vendors and line builders. DOORS, Polarion, Confluence, Jira, SAP QM, MES and the line IT stay what they are: your systems.
Each of our nine project offerings can be commissioned individually. The entry point is always a well-bounded slice with an existing data basis, on which the benefit can be measured before deciding on expansion.
The focus is on automotive – OEMs and tier 1, development, production and quality. The methods carry over just as well to medical technology, precision engineering, aerospace, shipbuilding and manufacturing.
Project offerings: development and engineering
Automated requirements engineering: RE Assistant. A checking rule set and an LLM assess requirements from DOORS, Polarion, codebeamer or Jama for testability, clarity and structural completeness – plus conflict candidates between related requirements. View the offering →
AI-supported test automation: Testomat. Test case proposals, traceability reconstruction, regression selection via change analysis, log evaluation, defect classification with a confidence threshold, reporting. View the offering →
Engineering Knowledge Assistant: EKA. Hybrid search with reranking across Confluence, DOORS, Jira and SharePoint – answers with source citations, refusal when the evidence is thin. View the offering →
Surrogate models for FEM/CFD. ML models predict simulation results in seconds: scalar quantities with stated uncertainty, result fields via geometry-aware operator and graph models. View the offering →
Automated engineering documentation. Test reports, release notes and ASPICE documents as auditable drafts from the source systems – with source references and automated evidence checking. View the offering →
Project offerings: production and quality
Visual quality inspection with computer vision. Anomaly detection on vision foundation models for rare defects, object detection for known defect classes, metrology for dimensional checks, drift monitoring in series production. View the offering →
Predictive maintenance and lifetime prediction. Two separately bookable modules: condition monitoring for production equipment and lifetime prediction for vehicle components with load spectra and censoring-capable reliability statistics. View the offering →
Root cause analysis in quality management. Statistical analysis of the process data around the defect, combined with research across past cases and quality records – hypotheses with evidence instead of a diagnosis. View the offering →
Project offerings: strategy and cross-cutting
AI Strategy & Transformation. Maturity assessment, use case prioritisation with business cases, data and AI governance along the EU AI Act in its Digital Omnibus version. View the offering →
What the offerings share
- People stay responsible: whether requirements review, test sign-off, document approval or root cause decision – the AI delivers justified proposals with sources, the decision stays with the engineer.
- Measured, not claimed: every offering names the metrics its success will show in. They are collected before the start and measured in the pilot, instead of being promised up front.
- Existing systems stay: all solutions dock onto the existing landscape – DOORS, Polarion, codebeamer, Jama, Confluence, Jira, SAP QM, MES, line IT.
- Data sovereignty with the customer: operation on-premises or in the approved cloud environment; the source systems' permissions apply inside the AI solutions too.
- Realistic scoping: initiatives start as a proof of concept or pilot with a small team and grow with proven benefit.
Reference projects
Detecting anomalies in industrial manufacturing. Machine learning models detect anomalous states in production data before they turn into scrap. View the use case →
Predictive maintenance in pneumatic systems. Damage and fault prediction on running mechanical systems based on sensor data. View the use case →
Predicting test bench durations for varying vehicle configurations. Regression models forecast the duration of test bench runs across heterogeneous vehicle configurations. View the use case →
From proof of concept to platform
Most AI initiatives do not fail at the model but at the transition into operation. We cover the entire chain ourselves – you need no second partner for industrialisation.
- Data Consulting – clarify the question and the data situation
- Data Lab – a prototype on your data
- Data Operations – production operation with MLOps
- Data Infrastructure – scalable cloud architecture, hosting in Germany and the EU
Collaboration and data security
We are certified to ISO/IEC 27001:2022 (TÜV Rheinland, reg. no. 01 153 2300149). Processing and hosting take place in Germany and the EU. We use client data solely for the agreed project and never to train our own models.
Engineering data is confidential: our solutions run on-premises or in the cloud environment you approve, and for language models the operating mode is your IT's decision – the architecture is unaffected by it. For the requirements of the EU AI Act we offer our AI Compliance Check.