Empresa:

JTA: The Data Scientists

Descrição da Função

Endeavoring to close the gap between Academy and Industry, we are opening Curricular Internship Positions for Master Students that want to develop their Master Thesis in a company environment.

All our Research Positions:

Are Paid Positions.

Are open for Hybrid or Remote Work.

Will emerge you in an Innovative & Young Culture.

Allow you to work on Real World Applications.

Focus on developing Cutting Edge Solutions.

Requirements:

• Background in Computer Science, Data Science, Mathematics, or a related field.
• Strong Python programming skills.
• Practical experience with Machine Learning libraries (e.g., PyTorch, TensorFlow) and image processing tools (e.g., OpenCV, scikit-image).
• Ability to work collaboratively in a multidisciplinary, fast-paced environment.
• Fluent in English, both written and spoken.

If you do not find below a project that you would like to work on, feel free to contact us, regardless. We may have other better suited opportunities for you. Similarly, we are always interested to listen if you have a project idea that you would like to implement collaboratively with us.

-

In machine vision environments-especially those deployed in industrial or high-throughput inspection systems-consistency in image quality is non-negotiable. Any sudden change in lens focus, lighting conditions, or unexpected objects entering the field of view (FOV) can compromise the reliability of downstream AI models and disrupt critical operations.

We are seeking a Master Thesis Student with a strong interest in Machine Learning and Computer Vision, who will explore anomaly detection techniques tailored to identifying such real-time issues in image acquisition systems. This thesis will be carried out as a curricular internship and will contribute to improving the operational resilience and reliability of visual inspection pipelines.

Main Objectives:

• Research state-of-the-art anomaly detection methods in imaging pipelines.
• Design and test machine learning approaches that flag capture-related degradations (e.g., blur, lighting drift, occlusions).
• Simulate and model real-world degradation scenarios to evaluate algorithm robustness.
• Integrate anomaly detection into a live image capture system.
• Document findings and provide a detailed final report, including implementation insights and suggestions for future work.

-

Object detection systems are increasingly being deployed in complex and uncontrolled environments, where objects can appear anywhere within the image. However, some detection pipelines exhibit positional bias, performing better when objects are centered or conform to training distributions.

We are looking for a Master Thesis Student to investigate how AI models can be made more robust to object location within an image. This internship will involve testing and improving deep learning architectures, data augmentation strategies, and training pipelines to ensure detection accuracy is consistent across the entire field of view.

Main Objectives:

• Benchmark state-of-the-art object detection models under varying spatial distributions of targets.
• Explore data augmentation, sampling, and architectural techniques to mitigate positional bias.
• Measure and visualize model sensitivity to object location across real and synthetic datasets.
• Propose and implement improvements to increase spatial generalization.
• Document findings and provide a detailed final report, including implementation insights and suggestions for future work.

-

At the intersection of art and technology lies a compelling challenge: determining the true authorship of a painting. In the field of fine art analysis, author attribution and forgery detection are complex tasks that require deep technical insight and sensitivity to visual nuance. Stylistic features, brushstroke patterns, pigment use, and compositional techniques can all serve as indicators of an artist's unique signature-but their accurate identification demands powerful and precise algorithms.

We are currently seeking a Master Thesis Student with a strong interest in Machine Learning and Computer Vision applied to the fine art domain. This internship includes a robust R&D component and offers the opportunity to collaborate with our expert team in developing models that assist in determining the likely author of a painting-or flagging potential forgeries-through digital analysis.

Main Objectives:

• Research and evaluate current SOTA techniques in visual style analysis, image classification, and artist-specific pattern detection.
• Design and train deep learning models that can distinguish artists based on subtle stylistic features in paintings (e.g., brushstroke geometry, texture patterns, compositional structure).
• Benchmark multiple model architectures and feature extraction pipelines.
• Collaborate with experts in both machine learning and art history to refine model interpretability and validate results.
• Document findings and provide a detailed final report, including implementation insights and suggestions for future work.

-

In textbook scenarios, machine learning often deals with clean, uniform data-but in the real world, particularly in cultural heritage and fine art digitization, data is rarely pristine. The accurate analysis of fine art images is often challenged by variations in capture environments, scanner technologies, and lighting conditions, all of which can introduce inconsistency. This poses a major obstacle when detecting minute surface features such as craquelure, fading, or structural degradation-crucial indicators in conservation and restoration efforts.

We are seeking a Master Thesis Student passionate about applying Machine Learning and Computer Vision to the Fine Art domain. The candidate will work closely with our team during a curricular internship with a strong R&D focus. The project will leverage our infrastructure and expertise to develop and evaluate algorithmic methods for detecting subtle differences and signs of damage in high-resolution images of artworks-while also ensuring image consistency across capture settings.

Main Objectives:

• Collaborate with the team to research and develop state-of-the-art (SOTA) approaches for detecting fine-grained differences in artwork images (e.g., craquelure, discoloration, structural wear).
• Enhance image consistency algorithms to minimize the impact of scanner, lighting, and environmental differences during post-capture processing.
• Build upon and improve internal deep learning pipelines by integrating recent advancements from academia and industry.
• Benchmark various detection and consistency techniques on internal datasets.
• Document findings and provide a detailed final report, including implementation insights and suggestions for future work.

Observações

Porto (Portugal)