We are a (hopefully growing) tech team of 15 people logically split in DEV and AI: Our product consists of the main application that connects to additional services, including AI services, running on kubernetes as web applications or spark jobs on GKE.
The AI services start with Spark jobs that are continuously processing large amounts of data. Other jobs, such as automatic registrations and documents associations, use that data applying different NLP and other ML techniques. For automatic processing of scanned documents and other images we use deep learning and OCR for text extraction.
Generally speaking, all our algorithms and heuristics are built in a way that the bookkeepers and accountants are kept in the loop so we are continuously receiving feedback and improving the models.
This is our tech stack
- Data munging – mainly with Python
- NLP & Machine learning: Python, Keras, Tensorflow, Seldon for AI model deployment
- Kubernetes, Docker, Prometheus for monitoring
- GCP (Cloud Pub/Sub, Cloud SQL, Cloud Task Queue, GKE, Cloud Storage, Google Functions, GoogleVision)