The aim of this thesis is to extend the existing latency analysis of a psychomotor feedback engine within our existing MLOps pipeline [1] [2]. Building upon preliminary latency estimations, this thesis will focus on systematically evaluating each processing step in the pipeline, assessing both theoretical and practical contributions to the overall latency and throughput. By modeling and analyzing latency sources, the goal is to propose and validate optimization strategies that can improve real-time performance for sensor-based AI applications. A particular emphasis will be placed on the throughput of parallel data processing within the infrastructure to ensure timely and efficient feedback delivery.
Thesis Type |
|
Status |
Open |
Presentation room |
Seminar room I5 6202 |
Supervisor(s) |
Stefan Decker |
Advisor(s) |
Michal Slupczynski |
Contact |
slupczynski@dbis.rwth-aachen.de |
Background and Related Work:
Real-time data processing for AI-driven sensor systems, particularly in psychomotor applications, demands strict latency requirements to ensure meaningful feedback for end-users. The MILKI PSY MLOps pipeline, developed to deliver psychomotor feedback, has established a preliminary framework across its processing stages, but gaps remain in its latency optimization. Existing research in sensor-based AI and real-time systems has focused on both minimizing processing delays and enhancing parallel data throughput, often leveraging techniques such as batch processing, data caching, and distributed processing architectures.
Research on latency management in MLOps pipelines has highlighted the significance of efficient data orchestration and pipeline scheduling in reducing lag. However, studies specifically tailored to sensor-AI latency in psychomotor feedback applications are limited, making this thesis an opportunity to advance the understanding of latency bottlenecks in such systems and their impact on throughput.
Expected Contribution:
This thesis will contribute a refined approach to latency optimization in sensor-based AI infrastructures, specifically addressing challenges in real-time psychomotor feedback applications. By proposing latency-reduction techniques grounded in both theoretical and empirical findings, the work will enhance the MILKI PSY MLOps pipeline’s responsiveness, laying the groundwork for future research in latency-sensitive AI applications. The findings will be of value to fields where high-throughput, low-latency processing of sensor data is essential, such as real-time health monitoring, robotics, and interactive learning systems.
If you are interested in this thesis, a related topic or have additional questions, please do not hesitate to send a message to slupczynski@dbis.rwth-aachen.de
Please apply with a meaningful CV and a recent transcript of your academic performance.
Prerequisites:
- Knowledge of MLOps and AI infrastructure – Understanding of machine learning pipelines, MLOps frameworks, and data processing.
- Proficiency in distributed and parallel computing – Knowledge of parallel processing, distributed systems, and throughput optimization.
Nice to haves:
- Understanding of sensor-based applications – Background in sensor data handling and processing, particularly in real-time feedback contexts.
- Skills in data modeling and performance evaluation – Ability to construct latency models, run simulations, and analyze system performance metrics.
- Experience with latency analysis and real-time systems – Familiarity with techniques for latency calculation, monitoring, and optimization.
Blockchain4DatenMarktplatz.NRW
MILKI-PSY - Multimodales Immersives Lernen mit KI für Psychomotorische Fähigkeiten