My research interests lie at the intersection of robotics, visual–inertial estimation, probabilistic perception, sensor fusion, and dependable artificial intelligence. I am particularly interested in autonomous systems that must operate with incomplete observations, changing sensor quality, uncertain state estimates, and safety-critical constraints.

My long-term objective is to contribute to dependable autonomy: systems that can quantify uncertainty, detect degradation before failure becomes critical, and initiate appropriate protective or recovery actions.

A reproducible research framework for estimator introspection, calibrated failure prediction, domain-shift awareness, and protective navigation. SHIELD-VIO studies whether visual–inertial systems can recognize loss of reliability early enough to protect downstream autonomy before localization failure becomes safety critical.

Python · Visual–Inertial Odometry · Error-State EKF · Uncertainty Calibration · Failure Prediction · Runtime Safety

• Reproducibility: deterministic configurations, documented assumptions, tests, metrics, and generated artefacts.
• Uncertainty awareness: diagnostic scores, calibrated probabilities, covariance, and consistency measures should remain conceptually distinct.
• Failure-oriented evaluation: average accuracy should be complemented by degradation studies, warning-time analysis, stress tests, and explicit limitations.
• Responsible claims: synthetic validation and research prototypes should not be presented as hardware validation, certified safety, or formal guarantees.

I am developing a focused research direction around uncertainty-aware and failure-resilient autonomous systems, with emphasis on visual–inertial estimation, failure prediction, runtime monitoring, and safety-oriented decision support.

Open to PhD positions and research collaborations in Robotics, Autonomous Systems, Computer Vision, and trustworthy AI.