Custom software & AI for industry — from shopfloor to ERP.
MES integration, predictive maintenance, computer-vision QA, production planning, digital twin. We deliver software that fits your manufacturing reality — and a fixed-price offer only when the prototype works.
Five common pain points for industrial mid-market
ERP, MES, machine data and Excel are separated. Reports are clicked together manually.
Predictive-maintenance projects end as PoC because nobody manages integration into existing systems.
The remaining 30 % is your competitive advantage — and your daily pain.
Stored in folders. An audit takes a week of work.
But nobody states clearly who builds, who operates and who audits.
Six solutions we build for industrial mid-market
MES & ERP integration
Make data flow between shopfloor, MES and ERP. Java/Spring/Kafka as backbone, OPC-UA as machine connector.
AI in quality assurance
Computer vision for visual inspection, anomaly detection on sensor data, RAG for maintenance know-how.
Predictive maintenance
Sensor streams via Kafka, models in Python, alerts into your maintenance planning. Including model card for the EU AI Act.
Production planning & scheduling
Custom planning tools that fit your manufacturing reality — not your Excel macros.
Supply chain automation
Supplier onboarding, order confirmations, goods receipt automated via Camunda + document AI.
Digital twin applications
Collect, visualise and simulate plant data. Web-based dashboards with real real-time integration.
Tech stack for industry projects
Backend & eventing
Workflow & orchestration
Machine & sensor integration
AI / machine learning
Frontend & dashboards
Infrastructure
Four use cases in detail
Real-time MES-to-ERP integration
Production books manually into ERP with 24h delay. We build a Kafka pipeline with OPC-UA reader on edge, Spring service in cloud, REST API to SAP/Microsoft Dynamics/Proalpha. Bookings in under 5 seconds, reports always current.
Bookings < 5 s
Predictive maintenance that reaches production
PoC shows good prediction accuracy, but nobody manages integration. We deliver: TimescaleDB for sensor history, Python model as REST service, alerts into Camunda maintenance workflow, Next.js dashboard.
Downtime reduced
AI quality inspection with audit trail
Visual inspection manual, error-prone, not scalable. Computer-vision model on industrial PC, REST API with classification + confidence, persistence in PostgreSQL, audit log for auditors. EU AI Act: model card, data lineage, human oversight.
Audit prep from days to hours
Order and supplier workflows automated
Incoming invoices, delivery notes, order confirmations — all manual from PDFs. We build document AI (LangChain + OpenAI), Camunda for approvals, integration to SAP. Result: significantly less manual back-office work.
Back-office work reduced
Compliance & data sovereignty for industry software
EU AI Act (effective Aug 2026)
High-risk AI in industry (e.g. machine safety components) requires conformity assessment. We deliver model card, data lineage and audit logs as standard artefact.
NIS-2 / NISG 2026
From October 2026 many industrial mid-market companies are obliged. As software supplier we support your supply-chain compliance with documented security measures.
Machinery Directive / CE
Software intervening in safety functions falls under Machinery Directive. We work with your TÜV/CE responsible.
Data sovereignty
Your production know-how stays in EU data centres. On request on-premises operation or hybrid architecture.
Frequent questions from industry projects
Yes. Via official REST/SOAP/RFC APIs, plus iDoc, BAPI or database replication. We confirm the concrete path during discovery.
Yes. We combine OPC-UA readers (e.g. Eclipse Milo) with Modbus-TCP adapters, normalise data in Kafka and deliver a unified view in dashboards.
Yes. NDA by default, EU data centres, on-premises on request. No data goes to OpenAI or other US providers without your sign-off.
We do not quote prices upfront or as flat rates. Workshop, domain analysis and prototype are free of upfront cost. A binding fixed-price or sprint-based offer is provided only once scope, architecture and risks are clarified based on the prototype. This avoids gut-feel estimates.
Yes. Llama, Mistral, Phi or fine-tuned models on-premises. Alternatively Azure OpenAI in EU region with data residency.
The discovery artefacts (architecture, domain model, risk register) belong to you — even if you choose another vendor.
Ready for an industry initial conversation?
30 minutes are enough to check whether our discovery model fits your industry initiative.