• Developed PoC and presented demo to leadership on innovative projects with Generative AI• Applied machine learning focusing on content evaluation, customer engagement, content IP discovery by collaborating across product, marketing, economics, and FP&A teams• Generated static and dynamic content and customer-based features, built models and applied causal inference techniques to predict and infer next best title for each customer to improve their engagement KPI • Worked with FP&A and negotiation teams on acquiring/existing content• Wrote papers and presented at Amazon machine learning conference Technical

• Performed cohort analysis, RFM analysis, lifetime value (LTV) modeling, and built a churn model for Lyft to identify factors influencing user LTV and churn. • Made recommendations based on the models and designed an experiment to lift LTV and reduce churn.• Selected as one of the six finalists out of forty-four teams to present at Top Project Showcase.

Spent the last two years of my PhD research at the Flatiron Institute on projects using machines learning to extract features and perform predictive analytics on hydrodynamic simulations to broaden the impact of my research and expand my collaboration opportunities internationally.• Trained machine learning models to map simulated luminosities of galaxies to their properties in small volume but high-resolution hydrodynamics simulations and ran such models on large volume N-body simulations to make predictions on the luminosities of large sample of galaxies, bypassing radiative transfer modeling.• Identified features (galaxy properties) most important in driving the scatters observed in scaling relations observed in simulated galaxies.• Used PCA analysis to reveal optimal projection of dataset (properties of galaxies in simulations) and reveal the parameters responsible for variance in data.

Forged new collaborations with scientists at the Flatiron Institute in NYC and Europe (UK, Denmark, Italy) to work on my latest projects ideas using simulations to confront observations. • Developed software to generate catalogs for extracting and storing galaxies and halos in cosmological simulation and automate the process on supercomputers (>500 GB RAM).• Developed tools in Python, supporting MPI, to simulate observable line emission from early galaxies in cosmological hydrodynamics simulations based on radiative transfer calculations.• Wrote shell/SLURM scripts to run Python scripts on supercomputers (>500 CPU cores).

• Worked with a team of theorists from Italy and US to examine properties of early galaxies.• Developed tools in python to analyze adaptive-mesh-refinement (AMR)-based cosmological zoom-in simulation data on supercomputers. • Devised 3D clump-finding techniques to identify structures in simulation data to examine the properties of galaxies in the early Universe.• Generated movies and animations of the simulations for presentations at conferences. • Results published in Leung et al., 2019c.

• Built an end-to-end project for pet service business owners to identify key concepts and insights across unstructured, lengthy customer reviews on Yelp for customer retention and marketing.• Webscraped 1M reviews, performed ETL and extensive numerical and visual EDA, data cleaning, and NLP. Implemented a probabilistic-based topic model (LDA) and SOTA Transformer model (Google’s T5). Reduced manual review reading time by >50%.• Designed streamlit dashboard and deployed a containerized interactive web app on Amazon AWS EC2 with Docker.

My thesis work focuses on examining the star-forming interstellar medium gas properties of galaxies in the Early Universe by modeling observations and analyzing numerical simulations on supercomputers, resulted in 5 first-author papers and contributed to 5 non-first author papers. Collaborated with scientists from >10 countries. Disseminated research results by presenting at >20 conferences and workshops.• Built tools in Python to perform parametric lens modeling using Markov Chain Monte Carlo (MCMC) sampling approach to reconstruct galaxy images from lensed emission • Wrote Python codes with YAML file to parse XML output from lens modeling code • Wrote Python codes to analyze astronomical (imaging and spectroscopic) data from interferometric observations (in Fourier domain)• Wrote Python codes and shell scripts (using SLURM) to perform spectral energy distribution modeling (using MCMC sampling) on supercomputers to reveal dust properties of galaxies • Led and crafted 10 successful proposals as Principal Investigator to acquire astronomical data from highly competitive facilities, contributed to >30 successful proposals as Co-investigator.

I taught 4 undergraduate-level Astronomy classes. • Astro 1101: "From New Worlds to Black Holes" with Prof. Lisa Kaltenegger• Astro 1102: "Our Solar System" with Prof. Phillip Nicholson and Prof. Steven Squyres• Astro 1195: "Introduction to Observational Astronomy" with Prof. Gordon J. Stacey • Astro 2201: "The History of the Universe" (writing course) with Dr. Shami ChatterjeeDuties include: • Grade writing assignments and hold office hours for 40+ undergraduate students. • Design and lead lab sessions to teach concepts including spectroscopy and optics.• Grade weekly writing assignments and provide constructive feedback to students.

Duties include:• Modify the design of the mating board for the antenna feed using DraftSight.• Coordinate the construction process of the HERA dish prototype.• Measure the response function of the HERA prototype through observations using spectrum analyzers.• Write Python-GPIB codes to generate an automatic sequence with front-end instrument for data collection (https://github.com/astro313/gpib)• Write Python code to strategize observation schedule, perform signal processing and transformation between topocentric and galactic coordinates (https://github.com/astro313/dishpaper)