Brendan Lee Bryant

Title: Brendan Lee Bryant and His Patents in Mechanical-System AI

Introduction

Brendan Lee Bryant is an inventor based in Kirkland, Washington. His patent work focuses on applying machine learning and artificial intelligence to control mechanical systems. He is associated with Microsoft Technology Licensing, LLC, and has 2 patents.

Latest Patents

Bryant’s recent patents address ways to make AI-driven mechanical control more effective and reliable.

“Machine Teaching with Method of Moments for Mechanical Systems” describes a method for training machine-learning models to explore a broader range of setpoint values. The approach augments system states with a fluctuating delta derived from a fixed setpoint, then uses a method-of-moments computation to convert inferred delta values into absolute values. Those values support a regression equation that a mechanical system can use to determine setpoint actions from its input states.

“Stochasticity Mitigation in Deployed AI Agents” presents techniques for reducing unpredictable behavior in AI agents that control mechanical systems. It distinguishes short-term from long-term stochasticity and applies different strategies to each. These include adjusting the model’s reward function for longer-term behavior and using output margins, such as window averaging, clamps, or statistical process control bounds, to manage short-term variation.

Career Highlights

Bryant’s disclosed inventions connect machine-learning model training with practical control of mechanical systems. His work addresses challenges such as limited variation in training data, unpredictable agent outputs, and predicted setpoints that fall outside permitted ranges.

Collaborations

Bryant’s listed coworkers include Kence Anderson and Kingsuk Maitra. His patents are associated with Microsoft Technology Licensing, LLC.

Conclusion

Brendan Lee Bryant’s patents focus on improving how AI models and agents determine setpoints for mechanical systems, with an emphasis on broader exploration during training and more stable behavior after deployment.