Worked closely with stakeholder to build nlp-based (LDA, word2vec, tf-idf) models to identify customer taste profile and top trending modules, leading to personalized product recommender collections. I was responsible for the full cycle of a model from scoping, development, training, modeling outputs, deployment into production environment, model A | B test lead lift in import KPR’s such revenue and conversion rate of ~2% and 5%, respectively. Proposed customer segmentation based on historical behavior combined with propensity model scores to identify more efficient marketing campaigns. These insights led to the increase in margin, better bucket allocation. Disseminated analytical insights and presented Machine Learning models across multiple departments. This led to the implementation of the customer profile infrastructure by multiple stakeholders (marketing, growth). Experience with full cycle of a model from scoping, development, training, modeling outputs, deployment into production environment, while working closely with different stakeholders such as acquisition, marketing product and data engineering teams. Worked with stakeholders, backend engineers (Sagemaker Feature Store) to stage and deployed A | B tests. Performed deep dive analysis to extract further analytical insights to facilitate post-test decisions.

Developed (Feature Engineering) and optimized (LightGBM) LTV model to outperform existingstatistical model version. Provide data-driven (customer segments) insights to identify actions to increase customer LTV. These insights led to the development of marketing strategies and A | B testing (App installation promos second first order promos) that impacted the budget allocation and increased customer lifetime value (up to 5%).• Developed product recommenders to improve first order rate, re-order rate and customer engagement with features provided by the company. This recommender ultimately lifted the ATC conversion rate up to 5% as part of marketing campaigns.• Performed statistical analysis of single variable incremental tests to estimate net cost per acquisition (CPA) for different acquisition channels to improve overall customer acquisition efficiency (Reduce CPA). Results were presented to stakeholders to optimize customer acquisition roadmaps.

-Lead numerical projects in collaboration with experimentalists focused on the self-assembly from proteins and nanoparticles via DNA-hybridization.-Proposed and developed clustering models on numerical data to understand the distribution of DNA chains in organic-inorganic materials that allowed to calculate DNA Chain Entropy.-Used t-test to calculate statistical significance between mean values of thermodynamics variables.-Developed force field modules for GPU-CPU based Molecular Dynamics module (Hoomd) written in Cuda, C++ and python (wrapper). These modules were implemented for research studies by me and other colleagues that have lead to multiple publications and conference presentations.

-Store podcast information in PostGreSQL databases and developed API (Listen Notes API) connections to existing podcast search engines.-Performed data cleaning on podcast descriptions and user Tweet posting history ( lemmatization, bi-and-tri-grams) with python (gensim, spacy, nltk)- Implemented Natural Language Processsing (pandas, tf-idf) to determine cosine similarity between podcasts and user's Tweeter history and used statistical analysis to select tailor-made podcast for the user. -Used LDA yo calculaste the coherence of proposed podcast for validations and also to select keywords to find podcast in the Listen Notes search engine.-Deployed tweet bot using AWS (EC2).

- During my research studies as a graduate student I focus on the self-assembly from nanoparticles, where I implemented Markov Chains (Monte Carlo) and Molecular dynamics to elucidate the role of particle shape and enthalpic interactions