This project aims to design and develop a prototype data center architecture that collects, stores, compares, and evaluates the outputs of various artificial intelligence models operating onboard an aircraft. The study will seek answers to critical questions regarding AI models performing distinct tasks across various domains on aircraft, such as their real-world operational accuracy, whether their performance degrades over time, and which model proves to be more reliable under specific operational conditions.

Criteria and Expectations

• A strong interest in designing data center architectures capable of aggregating and storing the outputs of diverse AI models in a standardized format.
• The capability to develop an evaluation engine that compares model predictions against actual outcomes (ground truth).
• Establishing analytical frameworks to detect model drift and recognize performance degradations at an early stage.
• The ability to measure the validity of models' confidence levels through confidence calibration analysis.
• Demonstrated aptitude for developing a dashboard prototype that visualizes and presents all analysis results to the user.

Relevant Academic Departments

• Artificial Intelligence Engineering / Artificial Intelligence and Data Engineering
• Computer Engineering
• Software Engineering
• Data Science and Analytics