MSc: AI Internship - Online Unsupervised Representation Learning for Printheads Behavior Analysis

Are you an HBO/WO MSc student in AI/Data Science who is looking for an (internship) graduation assignment? Are you ready for the challenge to improve our data analysis and monitoring tools with state-of-the-art machine learning methodologies? We are looking for you!

Assignment

You will work on the development of an end-to-end machine learning solution for analyzing the performance behavior of CPP’s printheads.

Introduction

Printheads, the core component of the inkjet printing process, continuously generate large volumes of telemetry data describing their operational state, usage characteristics, and health indicators of individual droplet-forming nozzles. Quality engineers and service professionals at Canon Production Printing use extensive data streams to monitor printhead performance, investigate quality issues, and identify emerging failure mechanisms. Due to the scale and complexity of the available data, this analysis is often time-consuming and requires access to substantial domain expertise.

The goal, and the main challenge, of this project is to combine data-driven approaches with domain knowledge provided by experienced engineers into a human-in-the-loop system. This is to help engineers better understand printhead behaviour, identify groups of similar printheads, and detect potentially novel or unusual patterns.

How

You will investigate how contemporary dimensionality reduction techniques, in particular deep metric learning, can be used to learn meaningful representations of printhead behaviour from high-dimensional operational data. These representations should support tasks such as clustering, similarity search, visualization, and novelty detection.

An important aspect of the project is the incorporation of expert knowledge into the learning process. Engineers possess valuable intuition regarding similarities and differences between printhead behaviours, and one of the research challenges is determining how such knowledge can be captured and reflected in the resulting machine learning models.

The exact methodology will be developed and refined throughout the project based on literature research, experimental findings, and discussions with domain experts and supervisors. The project builds upon existing research into spatio-temporal pattern formation in printhead telemetry and nozzle-grid logging data within CPP. Students are encouraged to explore both algorithmic and practical aspects of the problem, including machine learning methodology, evaluation strategies, visualization techniques, and software design considerations.

What

Expected deliverables include:

• A research report or a design document describing the investigated approaches and findings.
• Experimental evaluation of the developed methodology with reproducible code and supporting documentation.
• Recommendations for practical application.

As an optional extension, the developed methodology may be integrated into a lightweight interactive application that allows engineers to explore printhead populations, behavioural patterns, and potential anomalies. Successful outcomes may contribute to the validation of research results that are currently being prepared for journal publication.

Your profile

• You are currently pursuing a degree in Data Science, Artificial Intelligence, Computer Science, Electrical Engineering, Mathematics, or a related discipline.
• You can communicate technical concepts clearly to both technical and non-technical audiences.
• You have solid Python programming skills for data analysis and visualization.
• Basic familiarity with machine learning concepts such as clustering, dimensionality reduction, or deep learning is required.
• Experience with Pandas, NumPy, Matplotlib, Plotly, or similar libraries is preferred; experience with deep learning frameworks such as PyTorch or TensorFlow is a plus.
• Experience with software development is beneficial but not required.
• Familiarity with Git, REST APIs, Flask, FastAPI, Django, or frontend technologies is advantageous for students interested in the application-development component.
• Starting in September 2026, you are available for a period of 5–9 months and can work at least 4 days per week.

What’s in it for you?

• A challenging assignment with skilled coaching
• Internship/Graduation compensation of €500,- per month
• Travel cost compensation if you don’t have an ‘OV-weekcard’
• The possibility to network with professionals inside and outside your field of expertise, thanks to our diversity of disciplines which you will work with