Pierce Mayadag

• Developed prompts to train AI models in the fields of computer programming, mathematics, physics, research, spatial reasoning, and image processing.
• Effectively identified areas of weakness and designed prompts to work at the edge of model capabilities
• Thoroughly reviewed responses, articulating errors and providing ideal corrections.
• Created and applied multidimensional grading rubrics, both to individual models and in comparative analysis.

• Developed and maintained cross-platform, multi-language GUI physics simulation software tools.
• Interfaced directly with customers and users to identify desired features and fixes for the software.
• Proposed, acquired funding for, and lead a 2-man project to refurbish the testing suite of our software.
• Co-wrote a white paper proposing the use of Graph Neural Networks for physics simulation problems.

• Vastly expanded the Java API for Google's TensorFlow software & maintained it through version transition.
• Integrated the expanded TensorFlow API into an visual graph operations software, making it possible to build, train,
  and test neural networks via GUI.
• Used neural networks to approximate an optimal control function for Super Mario Bros using raw screen data as input.
  Network achieved ~90% completion rate of the first level.