About Client

Mentemedical is a award winning biotech startup leveraging IoT and AI technologies to help hospitals optimise instrument inventory.

They use RFID tags to track the surgical instruments. The RFID tags are attached to the surgical instruments. The RFID tags are read by the RFID Readers which are deployed in operating rooms. The RFID Readers send the data to the cloud. The data is processed by the software running on the cloud. The software provides insights to the hospitals.

Problem Statement

The existing RFID Readers were based on Beaglebone. They were comparatively expensive.

Making changes to the existing RFID Readers was difficult. We had to create a custom image for the Beaglebone. We had to flash the custom image to the Beaglebone for each change. This was a time consuming process.

Our Solution

We decided to switch to Raspberry Pi. Raspberry Pi is cheaper than Beaglebone. It is also more popular. This makes it easier to find resources and support for Raspberry Pi.

To speed up the CI/CD process, we decided to make the readers part of the Kubernetes cluster. This allowed us to use the existing CI/CD pipeline to deploy the readers.

Just by building a docker image, we were able to deploy the new reader software to the readers. This was a significant improvement over the previous process.

Implementation

We wrote a bootstrap script to setup the Raspberry Pi. It would do the following:

  • Install Tailscale
  • Enable CGROUPS
  • Intall K3s as agent and connect it to the cluster

This script would be run one time whenver we’d buy a new Raspberry Pi.

We then wrote a Dagger Pipeline to build an image for the reader client. The pipeline would build the docker image and push it to our image registry.

We also created a DaemonSet to deploy the reader client to the Raspberry Pis. The DaemonSet would pull the docker image from our image registry and deploy it to the Raspberry Pis.

We also terraformized the whole process.

Results

We were able to successfully switch the RFID Readers of Mentemedical from Beaglebone to Raspberry Pi. The new software is more reliable and cost effective.

The development process is also faster. We can now make changes to the reader software and deploy it to the readers in minutes.