Features - QEDK/clarity GitHub Wiki

Accurate

Our model has over 13M parameters, utilizing a bi-directional LSTM model built with TensorFlow Keras and an embedding matrix from GloVe dataset's 6B vectors of 300 dimensions. It has been trained on a TPU with accuracy and speed in mind, and provides sentiment analysis for 13 classes of moods of an average paragraph in <1 second.

Containerized

All our deployments are containerized using Docker and images are used to deploy revisions, allowing for easy traffic management, A/B testing and scalability. Cloud SQL and Cloud Run instances on GCP are configured to handle hundreds of current requests when required and be highly fail-safe.

Seamless

The project is deployed automatically from our codebase on GitHub to GCP Cloud Run seamlessly.

Asynchronous

All our code is written with async in mind, allowing concurrency for the best performance with the most efficiency.