Completed a text annotation project for a client building AI training data, applying hierarchical citation labeling in Label Studio across a large batch of documents — with a focus on accuracy and consistency rather than raw speed.

The client needed reference and citation data extracted from text with precision: correctly identifying authors, years, sources, and volume/issue fields, while also catching secondary (nested) citations and duplicate spans — details that are easy to get right once, but hard to apply consistently across dozens of documents.

I worked directly in Label Studio, following the client's exact taxonomy rather than improvising my own interpretation. After learning the labeling system, I annotated each document methodically, then ran a self-QA pass comparing early and later documents to catch drift in how rules were applied. Where a case was genuinely ambiguous, I flagged it instead of guessing — since incorrect confident labeling is worse than an honest question.

Delivered the project on time with consistent labeling standards maintained across 20+ documents. The client approved payment and indicated interest in future collaboration on related image annotation work.