Building a microservice based product seems like the common way to go, but it comes with a price. Each microservice has baggage, like CPU/memory footprint, which can be painful for Python based microservices. So how can we reduce overheads without giving up on microservices? I’ll present a solution which bunches Python microservices in a configurable way during runtime, allowing freedom in deployment to balance between encapsulation and reduction of a system’s footprint. I’ll show how my team created this mechanism, decoupled from everyday development of our microservices logic and share how we reduced the microservice count down from ~50 to ~10 per node, without compromising functionality.