Sujay Nair

I’m a third year math and CS double major at Georgia Tech. I also work as a student researcher at the MIT Climate and Sustainability Consortium where I’m lucky to be advised by Dr. Evan Coleman, Prof. Sherrie Wang, and Prof. Elsa Olivetti. My research focuses on using machine learning to infer properties of soil, with applications in carbon capture and mineral discovery. Previously I’ve had the chance to work on applications of machine learning in sports, robotics, and astronomy, and have spent time at NASA JPL. Outside of work I enjoy pottery, chess, basketball and football.

Email  /  CV  /  Google Scholar

In this project we developed a new architecture for inferring soil properties such as carbon from spectra.

In this project we train a 1D Convolutional Neural Network to take as input artificially generated light curves mimicking TESS observations at a 2 min cadence and outputa list with predicted planetary parameters-namely, Rp/Rs, Period, a/Rs, and T-mid.Comparing to the current state of the art algorithm, Transit Least Squares (TLS),which does not rely on machine learning, our Rp/Rs mean absolute error is roughly the same. We are still tuning the model to reach TLS performance for Period.

This summarizes my research in the area of updating exoplanet transits, crowd-sourcing the exoplanet transit effort, and deep learning for exoplanet detection.

We explore the benefits of using recurrent neural networks for predicting the exis-tence of an exoplanet. We tested LSTM layers in our Convolutional Neural Network(CNN) versus our standard CNN with synthetically generated light curves with vary-ing amounts of noise.

We extend the work of transit detection by utilizing phase folded light curves to traina convolutional neural network to predict the existence of an exoplanet. We propose a scheme to generate synthetic data mimicking TESS observations from a single sector where the positive training samples are randomly generated transit data phasefolded at the correct period, and the negative training samples are folded at a randomincorrect period.

6 members of the East Bay Astronomical Society will be introduced to the ExoplanetTransit Interpretation Code (EXOTIC) and will be guided through a complete transitand reference star analysis

We analyze the mid-transit times and comparison stars for the exoplanet TOI 1780.01.

We analyzed the mid-transit times and comparison stars for the exoplanets HAT-P-37 b and Kepler-45 b

We analyzed the transit properties of TrES-5b using many comparison stars and explored its extremely dim host star.

Analysis of transit properties of 2 exoplanets with 7 total image reduction methods.

We analyzed 4 exoplanets and edited the Exoplanet Transit Interpretation Code to provide faster photometric analysis.

We investigated 14 physical systems from the Washington Double Star Catalogue. We used our measurements to assess the probability of gravitational relationships.

From analyzing mid-transit times for many exoplanets, observations from small (<1m)telescopes can increase observational efficiency of large observatories.

In this study, we studied the effects of different comparison stars on the corresponding light curve and worked with 6 photometric functions.