* Built and implemented machine translation Spanish-English system , utilized seq-to-seq model and attention techniques to improve BLUE score, researched and implemented pointer network to generate summarization based on bank transaction data.• Built and deployed image facial recognition pipeline on Docker, implemented Fast-RCNN and MTCNN for efficient object detection by utilizing RPN, researched and built FaceNet on facial detection improving triplet loss.• Devised methodology for calculating the website traffic that is attributable to credit card advertising, implemented contextual bandit UCB algorithm to optimize ads performance• Developed query understanding/rewriting to predict user intent, increased user click through rate by utilizing WordtoVec(CBOW) and improved ads ranking by matching user preference to financial product.• Defined metrics and using geo experiments to estimating causal effects of COVID-19 product features changes on user experience, analyzed power and modeled network effect to decide unit of randomization, designed ramping up strategy.• Built data pipelines that pre-aggregate users across experiments/treatment groups & pre-aggregate metric data

• Developed an interactive web page (HTML/javascript) for users to update preference and view recommended games for users.• Developed a web service using (Java servlet, REST API) to fetch game data from Steam API, cleaned and purified essential features.• Utilized MySQL/MongoDB to store user preference and game information. • Designed and developed a filtering and sorting algorithm to match similar games based on game styles.• Improved the precision by ordering games based on review score and user recommendation.• Deployed the application to Amazon EC2 for better scaling performance.• Tested the web service with 5000 Steam user data.