👋 Hey there! I'm Evgenii.
I’m software engineer based in Amsterdam, The Netherlands. This is a personal blog where I (infrequently) write about code and other topics of interest to me. I currently work on Position Tracking at Semrush.
I’m software engineer based in Amsterdam, The Netherlands. This is a personal blog where I (infrequently) write about code and other topics of interest to me. I currently work on Position Tracking at Semrush.
A process is sending a lot of data. You have its PID and the remote address. Which file is it sending?
Picture this scenario: you’re writing Python code that calls a library for ML inference in the middle. You notice the program runs slowly. You collect a profile using standard tools like perf, and end up seeing this picture:
My friend who works on high-frequency trading systems once told me a story that completely changed how I think about database internals:
Working with large-scale data at Semrush, I’ve often faced the classic problem: “Is this item in my massive dataset?” When you’re dealing with billions of entries, even the most optimized database queries can become a bottleneck. That’s where Bloom filters come in - a probabilistic data structure that changed how I think about membership testing.
Design patterns aren’t just academic exercises - they’re communication tools. When you tell someone “this is a factory”, they immediately understand not just the structure, but the intent behind your code. In machine learning and data science workflows, these patterns become even more crucial as we deal with complex data pipelines, model training, and production serving systems.