I am a Senior Researcher at the Princeton Visual AI Lab lead by Professor Olga Russakovsky. I am involved in hands-on research, mentoring graduate and undergraduate students, as well as raising research funds for the Visual AI Lab. 

My research interest can be summarized as reverse engineering the reverse engineers: 

What lies at the heart of challenges for wide-spread adaptation of and safety concerns about AI models is that they are the first engineered systems that are not built from ground-up but top-down, making it difficult to understand failure mechanisms and to establish trustworthiness. I believe therefore it is essential to study these models just as we study natural systems and learn to control them.

Neural networks learn to approximate target functions presented to them in the form of training dataset and loss functions; thereby reverse engineering a target task function from data. I am interested in understanding how and what properties of a dataset and a given training procedure inform this approximation and how one can manipulate and control this process. 

I have a broad background. I previously worked on:

• Computer Architecture Design, developing architecture design and compiler tools for energy efficient and fast execution of Machine Learning and other memory bound applications such as Graph and Sparse Linear Algebra workloads; working on Architecture Design for Quantum Computing Systems and developing compiler tools mapping software level instructions to physical device backends.

• Data Science and Statistical Modeling, collaborating with Britt Hadar on studying the impact of risk appetite on biased representation of population subgroups in higher education and STEM; and Quantitative Finance (High-frequency Algorithmic Trading).

• Computational Biophysics (Ph.D), Molecular Biology and Genetics (B.S.).

Here is my contact info and publications.