1. Leverage AI and statistics to create impactful metrics that drive strategic decision-making for Google Shopping.2. Develop and improve experimentation infrastructure for Google Shopping.

Optimizing content selection and marketing strategies at Prime Video with machine learning, causal inference, and a wide range of AWS cloud services. Building end-to-end ML solutions to solve complicated problems.1) Built a pipeline to causally infer the impact of visual impressions over streaming to help optimize user engagement. 2) Built a NLP pipeline to identify similar movie and TV based on synopsis and genres, for unreleased titles, using language embedding.

Working in the one of the most impactful teams in Amazon to select "Add to Cart" (BuyBox) offer on Amazon Details page. Our service is called billions of times every hour globally. My contributions are:1. Re-architectured model generation and launching process. As compared to the popular approach of using observational data to generate models and then testing with A/B experiments, I designed the logic to holistically connect these two components, generating models through automated experimentation using Noisy Bayesian Optimization. This has greatly increased the success rate of all feature launches from 50% to >80%.2. Designed and automated experimental long-term success metrics for offer selection at Amazon product details page, which impacts over 10 million dollars Gross Merchandise Sales annually.

Design and build efficient experimentation platform for offer ranking algorithms, using rigorous statistical inference (AB testing, crossover design and multi-armed bandits), causal inference and machine learning.

Predictive modeling in clinical analytics using statistics, machine learning, python and SAS.1. Performed customer segmentation based on propensities of mental disorders, using regression and clustering.2. Predictive modeling of mental diseases on insurance claims, using random forest, support vector machine and boosting.

• Led the development of statistical genetics software for computationally detecting DNA mutations causing disease in correlated samples and in conducting Machine Learning analyses on genetic epidemiological data. • Wrote four published (one in a top journal, Nature Biotechnology) statistical software packages while serving as Lead Code Developer and Statistical Methods Contributor. Software packages were licensed by 1,100+ labs. Two software packages received invention-disclosure filing and one received patent-filing. • Participated in six statistical genetics projects, analyzing terabytes of DNA-sequencing data. Deployed methods in these projects include: composite likelihood ratio test and permutation test; generalized linear models; generalized random effect models; Maximum likelihood inference using MCMC; Marginal Likelihood calculation using Hidden-Markov Models; Bayesian hierarchical modeling. Successfully completed all assigned projects in a timely manner and published three journal articles in high impact journals. • Awarded honor of “Outstanding Trainee in Cancer Prevention.”

• Developed VAAST software in collaboration with colleagues and start-up to detect genetic mutations causing human disease. The VAAST software became popular in statistical genetics field and was patented. Software was commercialized by Omicia, Inc.• Collaborated with the research group of Dr. Mario Capecchi (2007 Nobel Price Laureate) to build machine learning models on cancer. Succeeded in constructing a Support Vector Machine model of clear-cell sarcoma and published my findings in the journal Cancer Cell.