Demo project of my talk for GalsenAI Dakar community on How to Deploy and Scale AI (ML or DL) Models with Kubernetes.

You can find the slides of my talk at https://meissa-gai-presentation.netlify.app.

This repo contains a REST API built with Flask that expose a machine learning model that can predict salary based on years of experience, test score and interview score.

You have two ways in order to setup and run this project.

For manual installation, you need to have Python (version 3 is preferable) on your system. Then you can clone this repo and follow the steps below :

• Create a virtual environment with the command :

    python3 -m venv venv
  

• Activate the virtual environment with the command :

    . venv/bin/activate or source venv/bin/activate
  

• Install the necessary dependencies with the command :

    pip install -r requirements.txt
  

• Start the server with the command :

    python app.py
  

• Go to your browser at the following address :

    http://localhost:5000
  

• Or make a POST request at http://localhost:5000/api/predict with the following payload (change values if you want) :

    {
        "experience": 1.8,
        "test_score": 12,
        "interview_score": 10 
    }
  

If you have Docker on your system, you have two options after placing yourself in the directory of this project :

• First option :

  • Build your image with the name you want like this :

      docker build . -t salary-prediction-service
    

  • Run it with the following command :

      docker run -d --name salary-prediction-service -p 5000:5000 salary-prediction-service
    

• Second option :

  • Run the image that i already deploy on docker hub with the following command :

      docker run -d --name salary-prediction-model -p 5000:5000 papihack/galsenai-salary-prediction
    

For the deployment of this project on K8S, i use a tool named Kind in order to create a local kubernetes cluster. You can also use other alternatives like minikube if you want it too.

The manifest files are located in the k8s folder.

• First and foremost, you need to have or create a kubernetes cluster. After you installed kind on your system, you can create one with the following command :

    kind create cluster --config k8s/cluster-config.yaml --name my-cluster
  

The file k8s/cluster-config.yaml will allow you to have a cluster with 1 Master Node (Control Plane) and 2 Worker Nodes.

• Then you can apply the deployment and service manifest files with the command :

    kubectl apply -f k8s/service.yaml -f k8s/deployment.yaml
  

• Finally, since it is a NodePort type service, you can try to find the IP of one your nodes and the generated PORT of the service in order to try to access to the app.

You can get those infos by running the following commands :

    kubectl get svc -o wide

    kubectl get no -o wide

The url should look like at something like : http://node-ip:svc-port for the web page and http://node-ip:svc-port/api/predict for the API.

API Interaction	Web interface

Feel free to make a PR or report an issue 😃

Oh, one more thing, please do not forget to put a description when you make your PR 🙂

• M.B.C.M