The current opportunity is available for current Manchester Met students, to apply you must be based in the UK for the duration of the role.
This internship is ideal for a motivated student interested in Artificial Intelligence (AI), Machine Learning, and Cybersecurity. The project will focus on developing and evaluating an AI model that is resilient to adversarial attacks—malicious inputs designed to manipulate machine learning predictions.
The internship will provide practical experience in adversarial machine learning, secure AI development, and model robustness evaluation.
The successful student will:
-
Conduct a brief review of adversarial machine learning techniques
-
Develop a baseline machine or deep learning model using Python frameworks such as PyTorch or TensorFlow
-
Implement adversarial attacks (e.g. FGSM or similar techniques) to evaluate model vulnerabilities
-
Design and implement defence mechanisms such as adversarial training or input filtering
-
Evaluate and compare the robustness of baseline and defended models
-
Document the methodology, experiments, and results in a short report and presentation
Weekly Timeline and Deliverables
Week 1: Project orientation and literature review
Deliverable: Short literature summary
Week 2: Develop baseline AI model and evaluate performance
Deliverable: Working baseline model
Week 3: Implement adversarial attacks and test model vulnerabilities
Deliverable: Demonstration of attack impact
Week 4: Develop defence mechanisms (e.g. adversarial training)
Deliverable: Improved model with defence techniques
Week 5: Evaluate model robustness and analyse results
Deliverable: Experimental evaluation and comparison
Week 6: Prepare final documentation and presentation
Deliverable: Final report (5–8 pages) and short presentation
Details
Dates: 25 May 2026 - 17 July 2026
Hours per week: 10
Students will be expected to develop the following skills and experiences while carrying out this internship
-
Secure AI development, including practical experience designing and implementing models with robustness and security considerations
-
Deep learning model development and evaluation using tools such as Python, PyTorch, TensorFlow, or Scikit-learn
-
Experimental research and data analysis, including comparing model performance and interpreting robustness metrics
-
Technical communication and research reporting, including documenting methodologies and presenting findings clearly
Please Note - As part of the 2026 Student Internship offering, priority will be given to Level 5 & 6 students who have specified in the Careers Registration survey (this is completed during enrolment and cannot be changed at this time) that they have had no work experience in the last 12 months and would like some on campus work or a short work placement. However, applications are welcomed from all students meeting the criteria below.
Students are only eligible to undertake one internship under the 2026 Student Internship offering.