Search Data Platform

Transformed customer conversational transcripts into structured and unstructured data to provide actionable insights and built out conversational commerce platform using AI & Machine Learning on vast sets of conversational data assets for customers worldwide. • Led, rearchitected and developed the streaming services that aggregates the messages into conversations with intents to support incoming Avro datasets from upstream services to Elastic Search; worked closely with data science team and planned with two software engineers. • Supported Kafka upgrade by creating index-comparison scripts for data consistency, creating backfill scripts for backup during deployment and fixing deployment issues on Kubernetes pods. • Automated the manual testing process by building a framework to produce canned datasets; speeded up the testing procedures by eliminating cross-functional dependency.

• Built a client-based self-service system that configured and automated the process of creating Airflow ETL data pipelines; saved 90% of the engineering efforts on handling customer requests.• Improved and upgraded the Insights on-prem platform software; wrote performance tests, fixed performance issues and created Prometheus metrics and alerts. • Automated the healing process on stuck platform widgets (streams, analysis, NLU models, dashboards etc.); saved 50% of the engineering efforts on resolving stability issues.• Handled 10+ customer requests of onboarding new customers, creating or updating current data pipelines and features weekly.

Measured and improved the accuracy of the device state for all Smart Home devices, across the world, connected to Alexa. • Designed and developed the back-end polling system that schedules the queries of millions of the Smart Home devices from 30+ partners to gather sufficient, even and continuous state data to improve the validity, reliability and quality of the state data. • Defined the quantitative measures of state data quality for all Smart Home devices and built the component that calculates the state data quality scores; aggregated and published scores to CloudWatch to enable public and real-time access. • Designed and led the implementation of a scheduler service that measures and computes the latency of customer’s update request to improve the measurement of state data quality. • Implemented and maintained the log sweeping tool that automatically converts the service data to Redshift database and provides the data to the Business Intelligence team to make business decisions. • Created a daemon that automates the device health-checking process from partners’ cloud to collect the device state data and measure the state data quality; reduced 90% time of the process. • Cooperate with important pre-launch Alexa partners to improve the state data quality by reporting and fixing the issues with device state updates.• Communicate and collaborate with multiple teams during daily operations work and design meetings.

Created an intelligent recommendation engine for people who intend to buy houses in City of Madrid.• Scraped, integrated and trained the social and geospatial feature data into cluster models.• Developed a request-response Web App using Django and JavaScript to provide a front-end questionnaire of social economics preference and house feature and recommend the housing options based on the user input.

• Designed and implemented the featured delivery window that provides the delivery slot which offers the promotion code to encourage the customer to select the most optimal delivery slot from Prime Now point of view.• Displayed the featured delivery window on the Prime Now check out Web App and Mobile App using JSP, JavaScript, CSS and AJAX.

• Collected the sports forecasters’ predictions on college football from Daily Illini.• Analyzed predictive data by comparing accuracy rate of forecaster and the betting line under Clustering Analysis.• Built an interactive system that filters, clusters and visualizes college football predictions of past, current and future.• Presented in University of Illinois Undergraduate Research Symposium.

Held office hours, designed and graded programming assignments for Introduction to Data Mining (CS 412) and Introduction to Differential Equations (MATH 285)

• Developed an underground water-flow and infiltration model to predict water density profile in the soil.• Simulated and animated the infiltration process by performing Central Difference Approximation in water dynamics.• Assessed the model in real agricultural data and obtained a similarity of 64% to the probed soil moisture data.