Dev Choudhary

I’m an AI Data Trainer focused on delivering high-quality labeled datasets that directly support computer vision model training and deployment. With 2+ years of experience, I improve annotation quality while keeping workflows efficient across distributed remote teams.

I specialize in semantic segmentation, instance segmentation, bounding boxes, keypoint detection, and pixel-perfect masking. I’ve produced 5K+ production-ready segmentation masks and labeled 3K+ social media images for e-commerce recommendation and entity extraction datasets.

I strengthen model training outcomes through RLHF-style evaluation, response quality assessment, and prompt/instruction tuning support. I’ve created 2K+ pixel-perfect image and video prompts with strong first-pass quality, and I audit large annotation volumes using dashboard-based tracking and QA metrics.

Beyond labeling, I’m hands-on with QA auditing, consensus scoring, ground truth validation, and SOP development. I mentor junior annotators through feedback cycles and calibration sessions, authoring 50+ SOPs to reduce error rates and standardize labeling best practices.