About Me
I am a Ph.D. researcher at Northeastern University developing foundational methods for trustworthy and scalable artificial intelligence in dynamical systems. My work lies at the intersection of data-driven learning, optimization, and control theory, with a focus on formal verification, robustness, and safety guarantees for neural-network-based decision-making systems.
I develop theory, algorithms, and software for certifying learning-enabled systems operating in complex, safety-critical environments, with applications spanning autonomous systems and robotics.
Research Highlights
- Scalable Verification: Developed a scalable stability and robustness verification pipeline for learning-enabled autonomous systems, achieving up to ~\(10^4 \times\) speedup over state-of-the-art methods.
- Neural Network Bounding: Introduced novel global and local sector-bounds for Feedforward Neural Networks (FFNNs), enabling depth-agnostic and input-aware bounds with tighter certification than methods like CROWN.
- Open-Source Libraries: Released two Python and MATLAB libraries: nnsector computes these local and global sector bounds for FFNNs, and nnstab uses them to check the stability of neural feedback loops and to find a region of attraction. See the Software page.
- Tool Integration: Integrated verification tools (AutoLIRPA, JAX-Verify) with classical control analysis (Lyapunov methods, MPC) to build reproducible verification workflows.
I am currently working on my thesis: “Scalable Method for Stability and Robustness Verification of Systems with NN Feedback using Positivity Constraints”.
