At SuperAnnotate, I contribute to AI data training and annotation workflows, evaluate data quality, and apply task-specific guidelines.
At BELAY, I annotated 200+ images with CVAT and maintained 98%+ accuracy across computer vision projects. I also reviewed and validated 300+ text and image annotations.
I developed annotation decision trees for ambiguous cases, reducing inter-annotator disagreement by 25%. I also implemented a systematic error-detection process that reduced dataset rework by 40%.
For LLM training, I evaluated outputs against safety policies and flagged hallucinations and policy violations. I also designed and iterated on 200+ prompt-response pairs with engineers.

