Serverless Data Enrichment Platform With Machine Learning in AWS
For an automotive client, we developed a serverless data enrichment platform leveraging SageMaker and HuggingFace transformers to classify customer feedback for aftermarket. We implemented CI/CD pipelines with scheduled and event-triggered retraining and custom monitoring for data and models. We build a wrapper over SageMaker ModelRegistry for model management. Data and model drift are measured and based on these metrics analytics we trigger re-training for all or a part of the models. We designed the system to implement shadow testing in production.