University of Oulu Computer Science and Engineering Master's Thesis Guide: 521993S, 30 ECTS, Examiners, 521009S Maturity Test and Laturi

Quick answer: what is the University of Oulu Computer Science and Engineering thesis route?

The current University of Oulu Computer Science and Engineering Master’s programme leads to a Master of Science (Technology) degree. The programme is 120 ECTS over two years, and the current 2026-2027 Peppi programme object is 52560 / IMP2026CSE. The exact thesis is 521993S Master’s Thesis in Computer Engineering, 30 ECTS, Peppi object 8306, classified as Advanced Studies and graded 1-5/FAIL. The current thesis implementation is 521993S-3005. The same programme requires 521009S Computer Science and Engineering, The Maturity Test for Master’s Degree, 0 ECTS, assessed pass/fail, with current implementation 521009S-3005.

The programme has four study-option environments: Artificial Intelligence, Applied Computing, Computer Engineering and Cyber Security. They share the same standalone CSE thesis object, but thesis topics and methods can differ considerably. Current CSE instructions use Laturi, a supervisor and second examiner, CSE-specific eligibility rules, a published 10-42 score-to-grade mapping and Degree Programme Committee approval. Programme 52560 does not establish a separate mandatory credited thesis seminar.

1. Confirm that you are in standalone CSE, not Business Analytics CSE

The University of Oulu also has a Business Analytics Computer Science and Engineering path, but that is a different programme object with a different thesis code. Standalone CSE uses 52560 / IMP2026CSE and thesis 521993S. Business Analytics CSE uses 521029S. Do not copy the Business Analytics thesis object into the standalone programme simply because both are related to CSE and may share some staff or methods.

2. The exact current thesis is 521993S and it is 30 ECTS

Current programme 52560 places 521993S Master’s Thesis in Computer Engineering inside a 30 ECTS Master’s Thesis and Related Studies module. The exact course is 30 ECTS, Advanced Studies, and uses the 1-5/FAIL assessment scale. The current teaching languages are Finnish and English. Satu Tamminen is listed as the current person in charge on the course object.

3. The current implementation is 521993S-3005

The live realization record identifies 521993S-3005 for the current 2026-2027 cycle. Its official dates correspond to the academic year beginning in August 2026 and ending in July 2027. Use this implementation only for the current evidence period. Later curriculum cycles may change the realization even if the main thesis code remains similar.

4. The thesis normally belongs to the second MSc year

The exact course places the thesis in the second year of MSc studies and states that compulsory advanced studies should precede it, shown as 90 ECTS in the current course information. This does not mean every thesis must start on the same date, but it does show that the thesis is intended after substantial advanced-level preparation. Check your approved PSP and supervisor advice before assuming you are ready to launch the project.

5. The programme has four study-option environments

Current CSE offers Artificial Intelligence, Applied Computing, Computer Engineering and Cyber Security, and students can choose or combine areas according to the study structure. Thesis method should reflect the actual topic. An AI thesis may focus on prediction or vision, Applied Computing may emphasise human-centred systems, Computer Engineering may involve embedded or signal-processing systems, while Cyber Security may require threat modelling and empirical security evaluation.

6. The thesis is independent work under supervision

The exact 521993S course states that the thesis is carried out independently under supervisor guidance. Independence does not mean working without feedback. It means the student is responsible for defining and executing the work, while the supervisor helps ensure the content, method and scientific or engineering quality are appropriate. Keep a clear record of major decisions so changes in scope remain understandable later.

7. The Degree Programme Committee approves key thesis roles and content

Current course information states that the Degree Programme Committee approves the supervisor, reviewers and thesis content. This makes the thesis more than an informal agreement between student and company. Formal roles and topic approval must fit the programme process. If the project changes substantially, confirm whether the approved setup also needs to be updated.

8. Start the formal process in Laturi

The current CSE thesis route uses Laturi for starting the thesis, supervision, monitoring, evaluation and publication. Do not wait until the final manuscript to open the process. The research plan, supervisor setup and later assessment are part of the Laturi workflow. Starting correctly reduces administrative problems near graduation.

9. Use the kick-off discussion to define the work

Current CSE guidance includes a kick-off discussion covering the topic, schedule, supervision implementation and assessment criteria. Use it to define what the thesis will actually deliver. Agree the core research or engineering question, expected data or system access, feedback frequency, company constraints and what evidence will be needed to support the final claims.

10. The research plan should make the evaluation logic visible

The CSE thesis process requires a research plan in Laturi approved by the supervisor. A useful plan should state the problem, objectives, technical or empirical material, intended methods, evaluation criteria, expected results and major risks. If access to hardware, company logs or participant data is uncertain, write that dependency explicitly instead of treating it as guaranteed.

11. The thesis must justify methods against the state of the art

The exact course learning outcomes require the student to justify methods or technological implementations in relation to the current state of the art. This means a thesis should not only explain what was built. It should show why the chosen method is reasonable compared with relevant alternatives and prior work. The comparison can be theoretical, empirical or engineering-based depending on the topic.

12. Testing and evaluation are central learning outcomes

The 521993S learning outcomes require the student to assess the functionality of the implemented solution using relevant testing and evaluation methods. A prototype that runs once is not enough evidence. Define what quality means for the project and select measurements accordingly, such as predictive performance, latency, energy use, reliability, security, usability, robustness or another justified criterion.

13. Compare results with the original goals

The thesis should compare final results against the goals set for the project and discuss their wider significance. Keep the original objectives visible throughout the work. If the result changes direction because the initial idea fails, explain that transparently. A negative or partial result can still be academically valuable when the evaluation is rigorous and the limitations are clearly interpreted.

14. Current CSE has no separate mandatory credited thesis seminar

Programme 52560 places only 521993S thesis and 521009S maturity test in the 30 ECTS thesis-related module. No separate mandatory credited thesis seminar is established there. If your research group or supervisor organises presentations or group meetings, follow those requirements, but do not invent extra seminar ECTS in the guide without a current programme object.

15. Current CSE defines supervisor eligibility

The current CSE instructions state that a thesis supervisor can be a CSE, EE or DCE professor or a postdoctoral researcher currently working at CSE, subject to the programme’s qualification conditions. Formal eligibility should be checked before thesis setup. A company mentor can support the project but does not automatically become the University supervisor.

16. A second examiner is part of the evaluation route

The current route includes a second examiner. Current instructions allow a CSE professor, a professor from another field or a postdoctoral researcher currently working at CSE, subject to the stated conditions. The supervisor and second examiner form the central evaluation pair in Laturi.

17. Check the senior-qualification condition early

Current CSE guidance requires either the supervisor or second examiner to meet the specified senior-qualification condition. A postdoc acting as supervisor must also have prior experience as a second examiner. Confirm the intended examiner combination early rather than discovering an eligibility problem after the manuscript is ready.

18. A technical supervisor may support external work

Current CSE instructions require the technical supervisor to have at least an MSc degree. This role is important when a thesis is carried out with a company or external organisation. Keep the technical and academic roles distinct. The technical supervisor can support implementation and domain access, while formal University assessment remains under the CSE academic process.

19. The supervisor and second examiner evaluate the thesis in Laturi

The supervisor and second examiner first discuss and agree on the evaluation. The supervisor writes the evaluation in Laturi, and the second examiner joins it. The exact course also describes the supervisor and second reviewer as the reviewers who provide the assessment proposal. Plan enough time for both reviewers to read the final version before the committee stage.

20. The current CSE score mapping is explicit

Current CSE instructions publish the score-to-grade mapping as 10-15 = 1, 16-21 = 2, 22-28 = 3, 29-35 = 4, and 36-42 = 5. Read the current evaluation criteria before writing the final thesis, not only after submission. It is easier to improve structure, scientific reasoning and evaluation while the work is still being developed.

21. Degree Programme Committee approval completes the formal grade process

The examiners prepare the assessment proposal, and the Degree Programme Committee approves the final assessment. The student’s upload date is therefore not the only milestone. The schedule should include supervisor review, second-examiner review, Laturi evaluation and committee handling before the intended graduation date.

22. Artificial Intelligence theses need a defensible validation design

The AI study option includes machine learning, machine vision and data mining. For predictive work, define what unseen case the model is expected to handle. Separate model development from final evaluation, choose metrics that match the problem and preserve enough information to reproduce preprocessing, training and testing. A high training score is not evidence of generalisation.

23. Data preprocessing is part of the scientific method

Towards Data Mining explicitly covers data collection, combining sources, normalisation, transformations, missing or incorrect values and generalisability. These choices can materially alter results. Report important preprocessing steps and keep them reproducible. If cases are excluded, variables transformed or missing values imputed, explain the rule and rationale.

24. Keep train, validation and test roles distinct

Current methods teach train-test-validation, cross-validation and related approaches. Feature selection, threshold choice and hyperparameter tuning should not repeatedly use the final test data. If the final test set is meant to estimate unseen performance, keep it outside development decisions. Otherwise the reported performance can become optimistically biased.

25. Grouped data need grouped validation

Many CSE datasets contain repeated observations from the same person, device, company or environment. Random row splitting can place closely related data in both training and test sets. If the intended claim concerns new users, devices or organisations, the split should reflect that unit. Explain why the evaluation design represents the real deployment problem.

26. Time-dependent data need chronological validation when relevant

Prediction in sensor, network, transaction or operational data often involves time. If the goal is future prediction, do not let future information influence model development. A temporal holdout, rolling evaluation or another time-respecting design may be more realistic than a random split. The exact choice should follow the deployment scenario.

27. Machine-vision theses should document image provenance

Current Machine Vision studies cover acquisition, feature extraction, transformations, 3D reconstruction, recognition and deep-learning fundamentals. A vision thesis should record image source, capture conditions, labels, preprocessing, augmentation and train-test provenance where these affect results. If multiple images come from the same subject or scene, avoid leakage across evaluation partitions.

28. Deep-learning complexity is not evidence of thesis quality

Current Deep Learning studies include CNNs, RNNs, transformers, GANs, autoencoders and diffusion models. Use a complex model only when the problem and data justify it. Compare against sensible baselines, report training configuration and limitations, and explain why the architecture is appropriate. A larger model does not automatically provide a stronger academic contribution.

29. NLP theses need corpus and model provenance

Natural Language Processing and Text Mining includes retrieval, text categorisation, corpus inference and language tools. Document how the corpus was collected, which languages are present, how labels were created and how duplicates or leakage were controlled. If external language models or APIs are used, record the relevant model or service version when possible and describe how outputs were evaluated.

30. Big-data scale does not remove methodological limits

Big Data Processing covers data-intensive systems, batch and stream processing, privacy, security and analysis. Large datasets can improve some technical tasks, but they do not automatically guarantee representative samples, independence or valid conclusions. Define the population and data-generation process before generalising from a large number of rows.

31. Multi-modal work needs alignment between data sources

Multi-Modal Data Fusion combines heterogeneous sensors or data types using statistical, Bayesian and machine-learning approaches. If your thesis combines images, text, sensor streams or other sources, explain how observations are aligned and what happens when one source is missing or noisy. Keep provenance for each modality so errors can be traced.

32. IoT theses should evaluate the end-to-end pipeline

The current Internet of Things course emphasises designing and implementing an end-to-end IoT software pipeline. An IoT thesis may involve sensing, networking, processing and application logic. If the claim concerns overall system performance, measure the relevant stages rather than only one component. Latency, reliability, energy, connectivity and data quality may all affect the final result.

33. Distributed-systems claims need an explicit environment

Distributed Systems covers communication, consensus, replication, consistency, scalability, cloud and edge computing. Performance claims depend on hardware, network conditions, workload and configuration. Record these conditions. A system shown to scale in one laboratory setup should not be presented as universally scalable without evidence supporting the broader claim.

34. Computer Engineering theses may require implementation-level evidence

The Computer Engineering option includes embedded systems, signal processing and IoT. Current Signal Processing Systems studies cover fixed-point implementation, filters, transforms, adaptive methods and Matlab/Simulink modelling. If numerical representation, processor constraints or simulation settings affect results, document them. Hardware and software versions can be part of reproducibility.

35. Cyber Security theses need a defined threat model

Security Engineering focuses on analysing complex systems, identifying weaknesses and making informed security-design decisions. A security thesis should define the security goal, attacker capabilities, trust assumptions and protected assets. Without a threat model, a statement that a system is “secure” is usually too broad to evaluate.

36. Empirical cyber-security research should be reproducible

Empirical Research in Cyber Security explicitly teaches replication of experiments and evaluation of scientific methods. Preserve tool versions, configurations, datasets and scripts when possible. Explain which parts of an experiment are replicated and which are modified. A security finding is stronger when another researcher can understand how the evidence was generated.

37. Human-centred studies need participant and task design

Applied Computing can involve human-centred systems, and current UX/Usability Evaluation studies cover participant selection, scenarios, methods, metrics, field/lab execution and analysis. If people participate in the thesis, explain who they are, why they fit the research question, what tasks they complete and what measures support the conclusions.

38. AI ethics, privacy and legislation can affect technical design

Current AI Ethics, Privacy and Legislation studies address ethical and legal conditions in AI development and deployment. If a system makes or supports decisions about people, consider data rights, fairness, explainability, potential harms and lawful processing. These issues should influence the design and evaluation, not appear only in a short final paragraph.

39. Personal-data planning must happen before processing

University data-protection guidance requires the personal-data lifecycle to be documented before processing begins. Address data protection and information security, perform risk assessment and consider a DPIA where applicable. Give research participants the required privacy information and collect only data necessary for the academic objective.

40. Pseudonymised data may still be personal data

Replacing names with codes does not automatically make data anonymous. If a key exists or individuals can reasonably be re-identified using other attributes, the data remain personal data. This can apply to sensor logs, device IDs, images, voice, locations, employee data and behavioural records. Keep re-identification material separate and restrict access.

41. Ethics review depends on the actual project

Not every CSE thesis requires formal ethics-committee review. Human participants, sensitive information, interventions or other applicable conditions may trigger preliminary ethical assessment. The Human Sciences Ethics Committee handles relevant non-medical human-sciences requests. Classify the project early with the supervisor so review, if required, occurs before the relevant research activity.

42. Build data management and research integrity into the workflow

Responsible-research guidance expects planning for collection, storage, access, sharing, preservation and reproducibility. Record ownership and access rights early, especially for company or licensed data. Research integrity also applies to code, external repositories, datasets and AI-assisted work. Keep attribution and AI-use reporting consistent with current University instructions.

43. The exact maturity test is 521009S, 0 ECTS

Programme 52560 requires 521009S Computer Science and Engineering, The Maturity Test for Master’s Degree, carrying 0 ECTS and assessed PASS/FAIL. Current implementation is 521009S-3005. The test is evaluated and approved by the thesis supervisor and can be completed when the thesis is complete or being finalised.

44. Current CSE guidance uses E-exam / Examinarium, while the exact course retains the written-event specification

The exact 521009S course description still describes a controlled written event on a topic provided by the thesis supervisor and gives an indicative length of about three handwritten pages or 450-600 words. Current CSE faculty guidance is more operational: the supervisor creates the maturity test as an E-exam, the student completes it in Examinarium, and the supervisor grades and locks it through the Exam system; the student must also register for the maturity test in Peppi. Treat the course description as the content/length boundary and the current CSE page as the delivery workflow, then confirm the active implementation with the supervisor before sitting the test. Follow the current individual language rule rather than assuming English for every international student.

45. Laturi submission must be public-safe

The exact thesis course requires a PDF/A copy in Laturi for assessment and archiving. The final thesis must not contain secret trade or professional information. If the project uses confidential company material, decide in advance what can appear in the public thesis and what must remain outside it. After approval, archiving and OuluREPO visibility follow the current permission rules.

46. Plan graduation separately from thesis writing

After required studies, thesis and maturity test are complete, degree application proceeds through Peppi’s graduation service. Current CSE guidance also requires an up-to-date approved PSP. Use the live graduation timetable when planning the final date because committee and application deadlines can change. A completed manuscript alone does not guarantee immediate graduation.

47. A practical 12-step CSE thesis workflow

1. Confirm programme 52560 / IMP2026CSE. 2. Confirm 521993S / 30 ECTS as the thesis object. 3. Find an eligible supervisor and confirm second-examiner requirements. 4. Complete the kick-off discussion. 5. Start the thesis in Laturi and submit the research plan. 6. Resolve data access, confidentiality, personal-data and ethics requirements before processing. 7. Build the system, experiment or analysis with reproducible methods. 8. Use validation appropriate to users, devices, time, systems or threat models. 9. Write the full thesis and compare results with goals and state of the art. 10. Complete supervisor and second-examiner evaluation and Degree Programme Committee handling. 11. Complete 521009S / 0 ECTS maturity test. 12. Finalise Laturi and apply for the degree through Peppi.

48. Final pre-submission checklist

Verify that 521993S remains the current 30 ECTS standalone CSE thesis, implementation 521993S-3005 is still relevant to your academic cycle, and 521009S remains the current 0 ECTS maturity test. Confirm supervisor and second-examiner eligibility, research-plan approval, current CSE evaluation criteria, data and ethics permissions, reproducible testing, public-safe company material, PDF/A Laturi submission, maturity-test language and live graduation deadlines. Finally, recheck the curriculum boundary: the University is transitioning to the 2027-2030 curriculum, so current 2026-2027 details should be freshly verified before they are treated as final rules for autumn-2027 study arrangements.