Fetchit
Computer Code Programming
Fine-tuningI worked on a data labeling project focused on financial Python code, things like trading logic, risk calculation scripts, and data ingestion pipelines. My job was to annotate and fine-tune training data, which meant reading code carefully enough to understand what it was actually doing before labeling it. I used Docker to keep the environment stable across the workflow. When you're dealing with financial code, inconsistencies in how tools run can quietly corrupt labels, so containerization wasn't optional; it just made sense. The Python side covered preprocessing raw code samples, cleaning them up, and building small scripts to automate parts of the labeling pipeline. A lot of the work came down to judgment calls: financial code can look similar on the surface but mean very different things depending on context, so accuracy mattered more than speed.