Nauto Uses ScyllaDB as a DB Layer for ML Algorithms

“I’ve worked with BigTable, Redis, Postgres and DynamoDB. Compared to all of these ScyllaDB provides very low latencies and has been inexpensive and easy to deploy.” – Rohit Saboo, ML Infrastructure Lead, Nauto. [Read Rohit's blog] 

Transcript

Nauto makes a device which we retrofit inside a car.  And this device has an internal camera and an external camera and sensors such as scopes, accelerometers, and so on. Fleets can use that to know which drivers are driving well, which drivers are not driving well, as well as to coach the drivers.

Some of the hard problems that we're trying to solve really span the space. We have to think all the way from hardware, to things on device, to cloud, as well as Machine Learning on the cloud.

So specifically, for me, it is understanding who the driver is, and that involves things like facial recognition, as well as non-visual modalities, such as based on prior history, understanding patterns, who's driving the vehicle, who drives at what time and, and so on.

ScyllaDB has been very helpful for us to be able to solve this problem by eliminating  a whole patchwork of solutions that we had, including Redis, Elasticsearch.  [We also] reduced the load, for example, some of them such as Kafka and Postgres and so on.

In the past, I have worked with Bigtable and I worked with databases such as Redis and Postgres as well as DynamoDB. Now, compared to all of these, ScyllaDB provides us very low latencies. Also, it has been pretty cheap and pretty easy to deploy, which has been very helpful to us.