Six Months Stage | Tetra Pak Srl

pytorch logo

Tetra Pak © is a Swedish multinational company specializing in food packaging and processing solutions. Founded in 1951 by Dr. Ruben Rausing and Swedish engineer Erik Wallenberg.

Tetra Pak operates in over 160 countries and employs more than 25,000 people worldwide. The company’s mission is to make food safe and available everywhere, driven by a commitment to sustainability and innovation. Tetra Pak’s product portfolio includes packaging, filling machines, and processing equipment for dairy, beverages, cheese, ice cream, and prepared foods.

During my internship as an AI Engineer in the D&T department, I focused on adapting Visual Language Models for domain-specific data to improve the efficiency of the company’s quality control process by automating the test environment. I researched and implemented state-of-the-art models, such as CLIP and SigLip, and devised ways to adapt them to Tetra Pak’s specific use case.

At the end of my internship, I presented my findings and recommendations to the team, highlighting the potential benefits of integrating Visual Language Models into the company’s quality control process. My work was well-received, and the team is now considering implementing my suggestions in future projects.

Throughout the internship, I faced several challenges, from data preprocessing to model fine-tuning. I manually annotated a dataset of a thousand image-text pairs, which helped me understand the importance of data quality in machine learning projects. I also learned how to fine-tune models using transfer learning techniques, improving their performance on domain-specific data.

The approach was successful, achieving an accuracy of 85% on the test set for classifying deviations from nominal conditions, a significant improvement over the baseline model.

Related Projects

Car Lo

I worked with two students on an IoT project using Arduino Uno, IMU, and GPS to improve driving styles and reduce CO2 emissions, with a demo app showing results.

View Project

Aws Solution Architect

The course introduced cloud computing, teaching me to deploy and manage open-source applications on AWS with a focus on cost constraints and Quality of Service (QoS).

View Project

LocoBot

I fine-tuned Yolov5 on the Hagrid dataset to predict manual gestures for tracking and stopping, enhancing the robot's capabilities while retaining prior knowledge.

View Project

MMR Driverless

As team leader, I coordinated the sensor system development for an autonomous vehicle, implemented the Stanley controller for trajectory tracking, and optimized skidpad performance for a formula student project.

View Project

Ovarian Cancer Survival Prediction

The project uses Graph Neural Networks on cancer patient data to enhance prediction precision by integrating multiple data sources with attention layers.

View Project

Placement of IoT Microservices in Fog Computing Systems: A Comparison of Heuristics

This paper proposes heuristics for efficiently placing microservices on fog nodes to enhance IoT application performance. It aims to minimize response times and meet Service Level Agreements by processing data closer to the sources, addressing the complexity and heterogeneity of fog infrastructures.

View Project

Artificial Intelligence & Analytics Engineer

Full time artificial intelligence and analytics engineer consultant

View Project