· Developed a Software Release Hub web application and service using React, TypeScript, Kotlin, and gRPC to allow internal users to view and edit previous and current software versions running on Caper Cart, enabling easier modification and version tracking for all developers

· Contributed to, debugged, and developed code for all supporting aspects of OTA updates, including multiple React applications in a micro-frontend architecture, Kotlin microservices, and Python agents running on IoT devices to manage software updates, logging, and automatic resolution for common Kubernetes errors

· Created an extensive unit and integration testing framework for frontend components and user flow with Jest and Cypress, as well as backend service tests using JUnit and Mockito

· Wrote, proofread, and assisted in writing documentation covering software architecture, the role of different microservices, and vocabulary used within and adjacent to the IoT Platform used to manage fleets of Caper Carts

· Developed an Item Upload Portal tool using React, TypeScript, Kotlin, and gRPC to allow clients to upload item information to the Caper database

· Transformed and migrated a Docker-Compose workflow to Kubernetes to allow for more flexible device monitoring and remote software and firmware (OTA) updates on Computer Vision services in Caper Cart

· Authored technical documentation for developer workflow in Computer Vision services to improve comprehension and setup time

· Led the charge for consistent code convention and formatting in the organization to save time on code authoring and review

· Developed a hierarchical user profiling mechanism for product and category recommendation based on user behavior, implemented in TensorFlow, improving the naive guess by over 700%

· Constructed an end-to-end system for the analysis (topic modeling and sentiment analysis) of over 200,000 e-commerce reviews and social media comments for several well-known FMCG brand clients

· Trained and deployed a prediction model for automatic product classification for thousands of items, achieving >99% accuracy and saving hundreds of hours of manual data entry

· Performed ad-hoc analyses of different user segments (modified RFM) for consumer targeting projects