Personal PoC of scoring Machine Learning models with Rust.
The project is structured as follows:
-
docker-compose.yaml: defines the services (features, score, grafana, prometheus)
-
dockerfiles/: single docker file for rust services
-
prometheus/: configuration files for prometheus
-
grafana/: configuration files for grafana
-
services/: Rust code
- feature_server/: code for a feature store. This is useful for batch features with infrequent updates, where something like redis would be an expansive overkill.
Here we keep track of the next steps:
-
add scoring service
- choose model (in case create rust bindginds for such model when needed)
-
improve prometheus metrics