The best use of Lambda is to programmatically respond to an event that was triggered by an AWS service, and typically have the response go to another AWS servic

The best use of Lambda is to programmatically respond to an event that was triggered by an AWS service, and typically have the response go to another AWS service. Lambda is a poor choice for tasks outside of this domain. We used it with great success in our data pipeline. Payload data comes into an S3 bucket, Lambdas will transform that data into a standard format and pipe the result to another S3 bucket. The data then needs to be filtered, post processed, then sent to many different data sources owned by different teams (inside or outside of AWS). Each of these steps are also a Lambda. The benefits are that up to a thousand concurrent Lambdas can be running, so performance doesn't really degrade with service loads. We can handle requests from dozens of teams all over the world with a team of two developers. We almost never have an unexpected outage. And, most importantly to me, we can add new features very quickly. For example, when a team builds a model in Python against the data, other teams want access to said model, our team can port it and integrate it into the pipeline in a day or two.