Context
At VBET, I worked on an analytics product for iGaming affiliates and influencers. It calculated financial and operational metrics in a platform that began serving audiences much larger than originally expected.
How I worked
Before addressing performance, I reduced security and maintenance risks in the legacy API. I replaced unsafe queries, organized the codebase around Clean Architecture and dependency injection, and established tests and documentation to support the following changes.
I then addressed performance in stages. I used Go, goroutines, channels, and parallel queries to reduce the first commission-calculation stage from about seven to three minutes. Since the SQL Server was external and could not be changed, I designed an ETL with checkpoints, pre-calculated aggregates, and reconciliation, separating provisional data from consolidated data.
What I took from it
This experience consolidated how I make performance decisions: understand the actual constraint, accept the consistency appropriate to each use, and then choose the technology that addresses it.