Neelanjana Pal (PhD)
About
I completed my PhD in Electrical and Computer Engineering at Vanderbilt University, specializing in the exciting field of AI systems verification. Since 2019, I've been actively engaged as a graduate research assistant in the veriVITAL lab, led by Dr. Taylor T Johnson, at the Institute for Software Integrated Systems.
My research journey revolves around the quest for safe and verifiable AI systems. At the core of my work is the improvement of Neural Network Verification Tools, specifically focusing on enhancing NNV, a tool developed by the veriVITAL Lab. I'm passionate about studying the impact of input perturbations on time-series based Neural Networks and applying Formal Methods to assess and enhance their robustness.
My academic pursuit started with an undergraduate degree in Electrical Engineering from Jadavpur University, India, in 2015. Following graduation, I ventured into the industry, where I spent two valuable years as a Project Engineer at Rockwell Automation India Pvt. Ltd. This experience allowed me to refine my technical skills and contribute to real-world engineering projects.
Alongside my doctoral journey, I had the privilege of interning at Mathworks, a global leader in software development. My work there was integrated into the 2022b version of Matlab, showcasing my dedication to pushing the boundaries of technology and software verification with a focus on responsible AI.
I am deeply committed to bridging the gap between academic research and practical industry applications while ensuring that AI technologies are developed and deployed responsibly and safely. I aim to leverage my academic insights and industry experience to drive innovation in AI verification and promote ethical AI practices. Furthermore, I am always open to networking, collaborating, and exploring new opportunities to make a meaningful impact in the world of technology and responsible AI.
Top Skills
Professional Experience
MathWorks
Involved working with the Fourier Neural Operator (FNO) to forecast outcomes for both familiar and unforeseen datasets (even at higher resolutions than the training data), leveraging the FNO's zero-shot super-resolution (ZSSR) capability to achieve super-resolution without the need for high-resolution training data.
Engaged in developing adversarially robust neural networks against attacks by employing different forms of Fast Gradient Sign Method (FGSM) adversarial training, including Basic Iterative Method (BIM) and Projected Gradient Descent (PGD).
Explored the robustness verification problem with time-series data using two key strategies: (1) robust training and (2) formal verification. While robust training could enhance model resilience against adversarial inputs through techniques like the FGSM and BIM; formal verification could offer formal guarantees of model behavior under various input scenarios.
- Grace Hopper Celebration 2024: [Workshop]
- International Women Engineering Day [Workshop]
- MATLAB Expo 2024 [Workshop]
- Women in Machine Learning 2024 [Reviewer]
The Climate Reality Project
Vanderbilt University School of Engineering
Website: http://www.verivital.com/
Vanderbilt University
- Introductory Programming (mainly Matlab): Fall'17 & Spring'18 - Provided guidance for Matlab programming
- Digital Logic (Classroom and Lab): Spring'18, Summer'18, Summer'19 - Conducted weekly lab sessions, performed lab report evaluations, and assisted with homework and exam evaluations
- Circuits II (Lab): Summer'18, Fall'18, Spring'19 & Summer'19 - Wholly responsible for conducting lab sessions and evaluation of lab reports
World Economic Forum
Website: https://www.globalshapers.org/hubs/nashville-hub
Instagram: @nashvilleglobalshapers
MathWorks
Worked on adding features for Neural Network Verification, published in MATLAB 2022b. This work demonstrates the direct impact of research on industry-standard software tools.
Rockwell Automation
- Hardware Engineering: System designing, PLC requirements, hardware configuration for clients including Rajasthan Dairy, Colgate Palmolive, Hindustan Unilever
- Networking and Virtualization: Projects for HIKAL & MARS International Ltd.
- Software Engineering: Process flow development in FTView SE for Hindustan Unilever, Banas Kanpur Dairy
- International Training: 3-month training in Shanghai, China, working on "Integrated factory with multiple assembly lines" project
Education
Vanderbilt University
Vanderbilt University
Jadavpur University
Indian Institute of Science Education & Research (IISER), Kolkata
IIT Kharagpur
Key Certifications & Training
Technical Skills
Research & Publications
Research Areas: Formal Methods, Deep Neural Networks, Verification, Control Systems, Cyber-Physical Systems
PhD Dissertation: "Reachability-Based Robustness Verification of Deep Neural Networks with Emphasis on Safety-Critical Time-Series Applications" (2024)
Email: Verified email at mathworks.com
Selected Publications
Volunteer Work & Leadership
- Impact Officer: 2022-2023
- Curator: 2021-2022
- Membership Board Chair: 2020-2021