I am a Post-Doctoral researcher in CMU Cylab hosted by Dr. Corina Pasareanu. I graduated with a PhD from the School of Computer Science at Georgia Tech in December 2020. [rmangal@andrew.cmu.edu] [CV] [LinkedIn] [Google Scholar]

Research Interests:
I am interested in developing formal methods for analyzing the correctness and safety of software systems. In recent years, I have focused on Trustworthy Machine Learning , i.e., robustness and explainability analysis of machine learning components as well as formal safety analysis of ML-enabled systems. Deploying ML models in safety-critical applications like autonomous vehicles demands techniques for establishing trust in the models and I believe that formal methods provide a powerful set of tools to address this trust deficit.

In my past and ongoing research, I have leveraged my formal methods expertise for developing tools and techniques that (i) analyze the robustness of deep neural networks (DNNs) to adversarial perturbations of the data, (ii) automatically extract high-level descriptions explaining the internal behavior of DNNs, (iii) formally verify that closed-loop autonomous systems with DNN-based perception comply with safety specifications, (iv) repair DNNs to ensure their compliance with user-provided safety specifications, and (v) quantify the uncertainty of DNNs for risk-aware downstream decision-making and controller synthesis.

(* indicates equal contribution, (#) indicates alphabetical ordering)