Developed a workflow-based application to create and evaluate different business cases using Vue.js with Node.js for the frontend and Dropwizard for the backend.
Implemented token-based authentication and RESTful APIs for secure, scalable communication between the frontend and backend.
Added multiple charts for data visualization and manipulation using chartJS.
Dockerized application for portability and added deployment using Jenkins CI/CD.
Enhanced the SmartHome system's scalability by inverting the client/server roles to centralize house control to the server rather than individual smart houses.
Data Acquisition and Cleaning to prepare large-scale (~10,000 homes) raw data from the SmartHome database for the classification model-training.
Custom AB Testing implementation with 'house-based feature flags' to evaluate the performance of the new ML-based Auto-Light-Turn-on Model.
Implemented Docker-based Continuous Deployment using Jenkins with support for test reports generation and deployment rollback.
Performed systematic software testing implementations including SonarQube setup, Unit testing with JUnit, Static Analysis with
Checker Framework, Integration testing with Mockito, along with Jacoco coverage testing and PIT mutation testing.
Developed a social chatbot that was awarded 1st place in Alexa Prize and acknowledged in EMNLP published article.
Implemented and annotated a new dataset (~1500 utterances) of Alexa human-machine interaction dialogue scheme based on the Universal Dependency annotation scheme to improve ASR error management. Result in the project's dependency parser model reaching 85.05% on labeled dependency edges.
Implemented an internal Automated Model Evaluation process to monitor latency, response time, create charting to discover outliers data, and analyze model’s performance bottleneck.
Performed feature debugging and model fine-tuning by daily user testing.
