General Thesis Information

On this page you find all the information you need if you are interested in writing a Bachelor’s or Master’s thesis at DOS. Please read the content of this page carefully before inquiring about a topic.

Who can write a thesis with us and what are the prerequisites?

In general, any student with an interest and some experience in the group’s research areas is welcome to write a Bachelor’s or Master’s thesis with us. To ensure the successful completion of a thesis, we see the following requirements:

• Completion of at least a seminar or a project or a bachelor thesis at our chair.
• Good programming skills in one or more of the following languages: Java, Scala, Python, C, and Go. The specific necessary programming language typically depends on the particular research topic.
• Familiarity with common tools such as git version control, text editors / IDEs, and typesetting tools like Typst (recommended) or LaTeX.
• Motivation to work scientifically and basic academic writing skills.
• Ability to work independently.

How do you get a topic?

Topics are offered by our individual group members. You can find a couple of thesis announcements on our website, but since not all ideas make their way there, feel invited to also make yourself familiar with our current research areas (e.g. take a look at the current projects and recent publications) and then contact the respective researchers directly. Please, do not write to the entire group. In any case, please provide the following information so that we get a first idea which topics might fit you well:

• Areas of interest for the thesis
• Your academic background (e.g. list of completed modules, transcript of records)
• A CV containing experiences in relevant topics from projects or employments
• What is the general process of writing a thesis?

Once you have met with one of our researchers and have found a topic that is interesting to you, we ask you to write an exposé of three to four pages in which you summarize the problem motivation, the related work and your idea for addressing the problem as well as an approximate timeline for your thesis project. Understand the exposé as a first version of the introduction section of your thesis later. Furthermore, it is often also a good reminder of how hard writing can be and helps to not forget the bigger picture in the beginning of the thesis project. Once you have received feedback on the proposal and you and the supervising researcher both agree with the exposé, you can register the thesis with the examination office.

During the thesis project, the supervising researcher will be available to provide guidance and feedback. At the end, we expect at least 30 pages for a Bachelor’s thesis and at least 60 pages for a Master’s thesis (single column, with a separate title page, table of contents, list of references). There is no official template at TU, so feel free to use any template that meets our requirements.

Please also take a look at the Study and Examination Regulations for your program, which contain further information about duration, language, and other conditions for the successful completion of a thesis.

Available topics

In the following you find currently available theses at the Bachelor’s or Master’s level. However, most of the topics are scalable and can be either expanded or narrowed down after consultation with the staff member.

If you are interested in one of the topics described below or would like to propose a related topic, please contact the respective person.

We strongly welcome individual modifications of the described topics and are open for additional suggestions. Please contact lehre ∂ dos.tu-berlin.de for any questions and suggestions.

Bachelor

• Model Compression Techniques in LLMs
• Continual Learning in Federated Learning
• Data Valuation and Critical Learning Periods in Federated Learning
• Resource Configuration for Graph Algorithms in Distributed Dataflow Systems
• Adaptive Memory Management for Scientific Workflow Tasks
• Root cause analysis of errors to increase IT reliability
• Carbon-aware federated learning

Master

• Finetuning of Large Language Models (LLM)
• LLM Routing (Query Complexity Estimation)
• Model Compression Techniques in LLMs
• Continual Learning in Federated Learning
• Data Valuation and Critical Learning Periods in Federated Learning
• Continual learning for Open Set Recognition with Deep Generative Models
• Dynamically Adjusting Cluster Resource Allocations for Batch Data Processing
• Resource Management for Adaptive Memory Requirements of Scientific Workflows
• Advancing Log Anomaly Detection and Root Cause Analysis for Enhanced IT Reliability in AI Operations (AIOps)
• Carbon-aware federated learning

Ongoing Theses

• Dynamically Adjusting Cluster Resource Allocations for Time-constrained Batch Data Processing with Apache Spark
  [M] Roman Guttzeit advised by Jonathan Will since 2024-06
• Ansätze zur Gewährleitstung der Sicherheit von Patientendaten in IoMT Systemen
  [M] Daniel Yermakov advised by Ilja Behnke since 2024-05
• Finetunig Named Entity Recognition Models on Domain Specific Datasets
  [B] Frédéric Ndjiki-Nya advised by Thorsten Wittkopp since 2024-05
• Insights into Distributed System Failures: Location Anomalies in Log Data
  [B] Anis Ben Saada advised by Thorsten Wittkopp since 2024-05
• Bidding-Based Distributed Scheduling for Offloaded Real-Time Tasks
  [B] Jan Läpple advised by Ilja Behnke since 2024-04
• Enabling Federated Learning to Interact with Energy Systems
  [B] Paul Kilian advised by Philipp Wiesner since 2024-04
• Avoiding Load Peaks in Testing-as-a-Service Applications
  [B] Mikolaj Cankudis advised by Philipp Wiesner since 2024-04
• Comparative Evaluation of Profiling-based Cluster Resource Allocation Approaches for Batch Data Processing
  [B] Anton Liudchyk advised by Jonathan Will since 2024-04
• Efficient Resource Allocation for Distributed Dataflows using Contextual Performance Modeling
  [M] Marvin Kronsbein advised by Dominik Scheinert since 2024-03
• Enabling Federated Learning to Interact with Energy Systems
  [M] Ovidiu Tatar advised by Philipp Wiesner since 2024-03
• Optimizing Renewable Energy Integration in Data Centers Through Co-Simulation
  [B] Julius Irion advised by Philipp Wiesner since 2024-03
• On the Feasibility of Lightweight Profiling for Performance Estimation of Distributed Data Processing Workloads in the Cloud
  [B] William Anton Knöpp advised by Dominik Scheinert since 2024-02
• Automation of Kubernetes-Based Experiment Setups for Distributed Dataflow Systems
  [B] Julian Nic Hahn advised by Dominik Scheinert since 2024-02
• Adaptive Anomaly Detection in LogData: Investigating the Role of Predicted Anomalies in Continuous Learning
  [M] Oscar Heimrecht advised by Thorsten Wittkopp since 2024-02
• Improving Renewable Energy Utilization in Data Centers Through Probabilistic Computation Offloading
  [M] Gesche Gräfe advised by Philipp Wiesner since 2024-02
• Vergleichende Analyse der vertikalen und horizontalen Skalierung für Spark-Datenverarbeitungs-Workloads
  [B] Dogukan Canatan advised by Jonathan Will since 2024-02
• Evaluation of Batch Workload Characterization Techniques for Performance Modeling of Distributed Data Processing Systems
  [M] Alexander Guttenberger advised by Dominik Scheinert since 2024-01
• Improving Dynamic Memory Prediction for Scientific Workflows
  [B] Sven Hoferichter advised by Jonathan Bader since 2024-01
• Dynamic Memory Allocation for Large Scale Scientific Workflows in Kubernetes
  [M] Julian Marcel Tochman-Szewc advised by Jonathan Bader since 2023-12
• Towards Predicting Runtimes of Distributed Batch Data Processing via Lightweight Profiling
  [B] Alaa Alhaidar advised by Jonathan Will since 2023-12
• Efficiency of Different Retraining Strategies for Machine Learning Models for Log Anomaly Detection
  [B] Jonas Möller advised by Thorsten Wittkopp since 2023-10
• A Distributed Scheduler for Offloaded Real-Time Task in Self-Organized Wireless Networks
  [M] Kalin Iliev advised by Ilja Behnke since 2023-10