Frequently asked questions and answers
Here you will find helpful information on the most frequently asked questions about our services and software solutions

General Questions

Yes. Data are processed either on-site at the customer’s premises or, if agreed upon, transmitted securely using end-to-end encryption. Divis handles data strictly within the scope of the respective project. Divis is not a data broker and does not use customer data to train proprietary systems. We comply with all applicable legal security standards, and our software is developed entirely in-house. For more details, please refer to our Privacy Policy.

Divis leverages sensor data, anomaly detection, and real-time analytics to identify potential failures early and predict optimal maintenance windows. The ClearVu Suite helps reduce downtime, optimize spare parts inventory, and increase asset availability—from pilot phase to full-scale rollout.

Step 1: Data & Sensor Audit (signal types, sampling rates, historical data)
Step 2: Use Case & KPI definition (e.g., Mean Time Between Failures – MTBF, downtime)
Step 3: Feature Engineering & Model Development
Step 4: Pilot deployment on a selected machine or production line, including monitoring

Divis provides end-to-end consulting—from data acquisition to production-grade deployment.

By implementing standardization practices such as MLOps (Machine Learning Operations), CI/CD (Continuous Integration/Continuous Delivery), modular architectures, quality gates, security reviews, and clearly defined operational SLAs. Proven models and artifacts from the PoC phase are hardened and transitioned into scalable services. Divis supports the entire lifecycle—from architecture and operations to change management.

Key considerations include data protection (GDPR), audit trails, traceability, and non-discrimination (EU AI Act). Divis, ISO 9001 certified, advises clients on policy frameworks and documentation. The ClearVu Suite offers comprehensive ML documentation capabilities for parameters, decisions, and versioning—critical for audits and compliance

Through automated data pipelines (validation, cleansing, enrichment), data quality rules, and human-centered QA checks. Divis’ data scientists define metrics for completeness, consistency, and timeliness.

Typical outcomes include reduced downtime, increased output, lower defect rates, and reduced energy/material costs. Divis brings use cases and expertise from two decades of industrial projects (e.g., automotive, process industries, chemicals, consumer goods)—with clearly defined KPIs and business cases.

Quantum-inspired and hybrid algorithms support routing, scheduling, and network optimization across supply chains. Divis provides strategic guidance, prioritizes application areas, and co-pilots solution deployment with your team.

AgenticAI

AgenticAI refers to systems composed of autonomous AI agents that pursue defined goals, make decisions, and execute tasks based on policy frameworks. In industrial settings, they optimize processes such as maintenance planning, quality control, and inventory management using real-time data. Prerequisites include clean processes, available interfaces, and high-quality data. Divis begins with a process/data assessment and maps value-adding steps to suitable agents.

An AI agent is an autonomous software component that uses AI techniques to gather information, perform analysis, support decision-making, or execute actions, ranging from research and reporting to real-time production control and logistics. Through the ClearVu Suite’s Python SDK, Divis provides agent modules for anomaly/pattern detection, AutoML (Automated Machine Learning), and optimization.

Success depends on data quality, a well-defined business case, interdisciplinary teams (domain experts, OT, IT, data science), MLOps maturity, security & compliance, and effective change management. Divis supports clients from initial workshops to international rollouts.

Through cloud deployment, multi-tenancy, internationalization/localization, and consideration of data residency and compliance requirements. Divis advises on scalable architectures for clients operating across multiple countries and regions (DACH/EU and beyond).

Technologies

Machine Learning (ML) leverages data to improve predictions and decision-making, the more high-quality data available, the more accurate the models become. In industrial environments, this translates to higher Overall Equipment Effectiveness (OEE), reduced scrap rates, and more stable processes. Using AutoML within the ClearVu Suite, divis intelligent solutions GmbH automates model selection and optimization, seamlessly integrating results into existing manufacturing IT systems (On-Premises, Cloud, Edge).

Deep Learning is a subfield of ML that utilizes multi-layered neural networks to learn highly complex patterns from large datasets (e.g., image, audio, or time-series recognition). Divis applies Deep Learning models selectively where they demonstrably add value, such as in visual quality inspection or Multisensor Fusion.

Supervised learning trains models on labeled data (e.g., good/bad, numerical values) to make targeted predictions or classifications. Unsupervised learning identifies patterns and structures in unlabeled data (e.g., clustering, anomaly detection), uncovering insights in process or sensor data. Divis selects the appropriate method based on the task, data availability, and business objectives.

Through real-time monitoring (e.g., input/prediction drift, latency, error rates), alerting mechanisms, automated retraining pipelines, canary releases, and rollback strategies. This ensures long-term model robustness, stable performance, and rapid adaptation to process changes.

Multi-Objective Optimization (MOO) simultaneously optimizes multiple, often conflicting goals (e.g., cost, quality, energy consumption, emissions). The result is a set of Pareto-optimal solutions that enable balanced decision-making, ideal for production and supply chain scenarios. Divis defines objectives and constraints, models them using ClearVu AutoML, and executes data-driven optimization.

Using systematic search strategies (grid/random), proven evolutionary algorithms, early stopping, and feature selection, models become more stable and accurate. The ClearVu Suite provides scalable pipelines for this purpose, proven in use cases such as production ramp-ups, calibrations, and predictive maintenance forecasting.

Quantum algorithms operate using qubits and quantum phenomena (e.g., superposition, entanglement), offering advantages for specific problem classes such as optimization, simulation, and cryptography. Classical AI/ML remains dominant for most production tasks; quantum computing complements these in hybrid approaches. Divis evaluates suitability, data, and objectives through structured workshops.

Software

Via the ClearVu GUI, an Excel add-in, and a Python package/SDK, divis embeds ML functionalities into existing systems. Typical interfaces include REST/HTTPS APIs, OPC UA, MQTT, and file services (CSV/Parquet). Deployment is flexible, On-Premises, Cloud, or Edge.

Using cross-validation, clearly defined performance metrics, and benchmark/holdout tests. The ClearVu Suite’s Algorithm Tournament compares candidate models reproducibly, provides documentation (Model Cards), and supports A/B rollouts, ensuring model quality is validated before production deployment.

To ensure robust interoperability, divis supports standard formats such as CSV, JSON, and XML, along with common industrial protocols (e.g., OPC UA, MQTT). Structured and versioned data provisioning facilitates AutoML pipelines, reproducibility, and compliance-ready documentation.

ML for predictive quality relies on structured historical data, labeled outcomes, and relevant process variables. Key metrics include MAE, F1 score, and Recall@Top N. The ClearVu Suite’s Algorithm Tournament ensures reproducible model comparison, documentation, and A/B rollout support, validating model performance prior to production use.

Modeling based on data to uncover underlying relationships is a recurring task. In practice, this is often done using Automated Machine Learning (AutoML), which includes a wide range of modeling techniques such as neural networks, random forests, support vector machines, and radial basis functions. While many of these methods yield satisfactory results with default settings, significant improvements can be achieved by optimizing the parameters that govern the training process, known as hyperparameters. The goal of hyperparameter optimization is to minimize error within a cross-validation framework, ensuring strong generalization capabilities. Once optimization is complete, a set of tuned hyperparameters is available to initiate the final training process.

Through visualizations, charts, and model transparency (data and decision pathways), users receive interpretable results. ClearVu provides explanations and rationales for each model version, enhancing trust and accelerating approval processes.

Using the ClearVu GUI, Excel add-in, and Python SDK, divis integrates AI functionalities seamlessly into existing IT/OT environments, including APIs and monitoring capabilities.

Have more questions?

We’re here to help!

Telephone

+49 231 97 00 340

E-Mail

contact@divis-gmbh.com

Have more questions?

We’re here to help!

Telephone

+49 231 97 00 340

E-Mail

contact@divis-gmbh.com