Machine Learning und Computer Vision

Alan Stimac

Senior Project Manager

+41.. Show number

Expert knowledge for your project

Alan Stimac is a senior project manager at soxes with a technical background and a master’s degree in mechanical engineering from University of Zagreb. He reliably guides customers through complex IT projects and ensures that solutions are not only technically sound, but also remain stable, understandable, and maintainable in operation.

We develop AI models that work in real-world processes

Many companies already have valuable data: images from inspection processes, measurement values from equipment, sensor data from products, or historical process data. However, the real value only emerges when this data is transformed into reliable models, clear analyses, or automated inspections.

We assess whether your data is suitable for this purpose, develop appropriate machine learning or computer vision models, and integrate them into existing applications, systems, or processes.

Machine Learning and Computer Vision with soxes

1. Your question	2. Our response
What does soxes offer?	Feasibility studies, data analysis, model development, computer vision solutions, quality control, measurement data analysis, model integration, and MLOps.
Who is this relevant for?	For companies that want to use images, measurement values, sensor data, or process data to support inspections, forecasts, evaluations, or decisions.
What problems do we solve?	Manual visual inspections, unused measurement data, hard-to-detect patterns, inconsistent quality, a lack of predictive analytics, and machine learning prototypes that never make it into production.
What makes soxes different?	We combine data science with software development and systems integration.
What references are available?	GRS Gemresearch Swisslab AG, Emmi Group AG, Electrolux, BBT, Stadler, Ergodent

Here's what we offer in the areas of machine learning and computer vision

We are assessing the feasibility

We assess whether data, images, measurement values, or sensor data are suitable for machine learning or computer vision.
We process data

Data is structured, cleaned, and analyzed so that models can produce reliable results.
We develop ML models

Models identify patterns, classify data, generate forecasts, or assess technical conditions.
We develop computer vision solutions

Images are analyzed, objects are detected, defects are identified, or quality characteristics are checked.
We train and validate models

Results are tested, accuracy is evaluated, and practical applicability is thoroughly assessed.
We analyze measurement data

Technical data, sensor data, and process values are made understandable and usable for decision-making.
We integrate models into software

Models are integrated into existing applications, processes, systems, or production environments.
We support the pilot project

Monitoring, versioning, updating, and further development are taken into account from the very beginning.

Common issues we encounter

1. Data is available but unused
Images, measurement values, or process data are collected but not systematically analyzed.

2. Data quality is unclear
There is a lack of structure, labels, reference values, or clean data sets for reliable models.

3. Internal technical expertise is lacking
Data science, model training, validation, and integration require experience that is often not permanently available within the company.

4. Manual checks take too much time
Visual inspections, quality checks, or evaluations tie up specialists and remain prone to errors.

5. Errors are detected too late
Deviations, patterns, or anomalies are difficult to spot in large data sets.

6. A prototype doesn’t work in everyday use
Initial models deliver results but cannot be stably integrated into real-world processes.

7. Accuracy is insufficient
Models appear promising, but their accuracy, stability, or traceability are insufficient for everyday use.

8. Operations are not clearly defined
Without monitoring, versioning, and updates, a model loses quality over time.

References from machine learning and computer vision

1. Emmi

Emmi wanted to make packaging inspections more reliable and efficient. To achieve this, soxes developed a machine learning solution that identifies relevant features and supports quality inspection with data-driven insights. This makes deviations visible more quickly and improves the usability of inspection processes in the production environment.

2. Electrolux

Electrolux processes large volumes of service requests and repair data. soxes developed an AI-based repair configurator as a proof of concept. The solution analyzes error messages, product groups, and historical repair data to suggest suitable replacement parts. This makes technical data actionable, reduces support inquiries, and better supports repair processes.

3. GRS Gemresearch Lab AG

For the digital presentation of certified gemstones, soxes developed a native AR app with integrated computer vision. Since gemstones exhibit only subtle visual characteristics, standard solutions like ARKit and ARCore were insufficient for recognition. soxes therefore trained its own machine learning model for image analysis. This model identifies certified gemstones based on visual characteristics, retrieves relevant data via APIs from the new LIMS, and enables a stable display of relevant content in augmented reality.

Prototype or production ML solution?

Prototype	Productive ML solution
The data must first be reviewed	Data quality and the target vision have been clarified
Accuracy and usefulness remain to be seen	The model is intended for everyday use
The first results should become apparent soon	Integration into software or processes is necessary
Risks and costs must be assessed	Monitoring, operation, and further development are important
The use case needs to be reviewed by subject matter experts	The solution is designed to deliver consistent results

In 30 minutes, you'll know if your data is suitable for machine learning!

Do you have images, measurement values, sensor data, or process data and want to know if they can be turned into a useful solution?

In a brief conversation, we’ll determine what data you have, what use case is realistic, and what the next logical step would be.

Direct number

+41 55 253 00 53

Book your initial consultation now!

Digitalization and

Individual Software

Machine Learning und Computer Vision

Contents

We develop AI models that work in real-world processes

Here's what we offer in the areas of machine learning and computer vision

We are assessing the feasibility

We process data

We develop ML models

We develop computer vision solutions

We train and validate models

We analyze measurement data

We integrate models into software

We support the pilot project

Common issues we encounter

References from machine learning and computer vision

Prototype or production ML solution?

The data must first be reviewed

Data quality and the target vision have been clarified

Accuracy and usefulness remain to be seen

The model is intended for everyday use

The first results should become apparent soon

Integration into software or processes is necessary

Risks and costs must be assessed

Monitoring, operation, and further development are important

The use case needs to be reviewed by subject matter experts

The solution is designed to deliver consistent results

In 30 minutes, you'll know if your data is suitable for machine learning!

Trusted by Leading Brands

This might interest you

Do you have any questions? Would you like to find out more about our services?
We look forward to your enquiry.

Top Rated Service

Digitalization and

Individual Software

Machine Learning und Computer Vision

Contents

We develop AI models that work in real-world processes

Here's what we offer in the areas of machine learning and computer vision

We are assessing the feasibility

We process data

We develop ML models

We develop computer vision solutions

We train and validate models

We analyze measurement data

We integrate models into software

We support the pilot project

Common issues we encounter

References from machine learning and computer vision

Prototype or production ML solution?

The data must first be reviewed

Data quality and the target vision have been clarified

Accuracy and usefulness remain to be seen

The model is intended for everyday use

The first results should become apparent soon

Integration into software or processes is necessary

Risks and costs must be assessed

Monitoring, operation, and further development are important

The use case needs to be reviewed by subject matter experts

The solution is designed to deliver consistent results

In 30 minutes, you'll know if your data is suitable for machine learning!

Trusted by Leading Brands

This might interest you

Do you have any questions? Would you like to find out more about our services? We look forward to your enquiry.

Top Rated Service

Your request

Do you have any questions? Would you like to find out more about our services?
We look forward to your enquiry.