AI Safety, AI Planning, Robotics, Action-Model Learning, and Analysis of Abstractions.
My research focuses on safe and reliable behavior of AI agents. I am currently investigating the minimal set of requirements in an AI system that would enable a user to assess and understand the limits of its safe operability. In the past, I have worked in the area of bio-inspired robotics, speech processing, and internet of things.
Investigating the minimal set of interfaces in an AI system that would enable a user to assess and understand the limits of its safe operability.
Developed a framework which generates interrogation policies to derive the human interpretable model of an autonomous agent.
Reduced the STRIPS-model comparison to a planning problem which enabled identifying if two models are not same.
Modified FF-planner to take PDDL models, initial state, plan as input, and give the final state as output.
Co-designed the experiments showcasing possible use-case of an NSF project on training lay people to use adaptive AI systems which are safe, robust, and reliable.
Workload Prediction: Predicted workload on a storage volume based on I/O patterns for All Flash systems. This helped in optimizing data storage and management while keeping the nature of data private to the users.
File and Volume Clones: Developed features and managed issues related to file and volume clones on UNIX-based NetApp proprietary WAFL file system so as to improve the storage efficiency of enterprise data servers.
Co-developed a file system component which prevented data loss if files and volumes are cloned or modified when cloud backup is offline.
Portfolio Management Applications: Built applications for iPhone and iPad, and developed tools so as to analyze the user portfolio in an interactive way.
Cross-platform Applications: Cross-platform mobile application development of mobile applications supported on both Android and iOS.
Teaching Experience
Spring'22
Teaching Assistant for a graduate-level Artificial Intelligence Course at Arizona State University.