Tao R.

Key Skills

Software

Other

Sama

Top Subject Matter

Autonomous driving

smart city computer vision datasets (drone and fish-eye imagery)

Video segmentation for temporal pixel-level CV labeling

Top Data Types

Video

Text

Image

Top Task Types

Bounding Box

Segmentation

Prompt + Response Writing (SFT)

Mapping

Freelancer Overview

Senior Data Annotation Engineer & Project Lead — CV Annotation Projects. Core strengths include Other and Sama. Education includes Bachelor of Science, Xi'an University of Arts and Sciences (2014). AI-training focus includes data types such as Geospatial, Tiled Imagery, and Video and labeling workflows including Bounding Box, Segmentation, and Prompt + Response Writing (SFT).

Labeling Experience

Multimodal Large Model Data Governance Specialist (Contractor) — OVD/VLM Ground Truth Curation

OtherImageMapping

Curated multimodal datasets for open-vocabulary detection (OVD) and vision-language models requiring text-to-region ground truth mapping. Produced region-level supervision aligned with text queries to enable autonomous vision agent training. Automated pre-labeling steps for named entity recognition and relationship extraction to speed up data cleaning and reduce manual workload. • Text-to-region mapping ground truth for OVD/VLM training • Region supervision curation for multimodal alignment • NER and relationship extraction pre-labeling automation • Data cleaning cycle acceleration through local lightweight model deployment.

2024 - Present

Multimodal Large Model Data Governance Specialist (Contractor) — LLM Training Data

OtherTextPrompt + Response Writing (SFT)

Constructed premium SFT datasets by producing high-quality prompt-response dialog utterances for LLM training. Built RLHF preference annotations by collecting and structuring human feedback signals for model alignment. Evaluated and cross-verified multi-turn responses using rigid criteria to ensure logical consistency, structure integrity, and alignment safety boundaries. • SFT prompt-response dialog data creation • RLHF preference annotation and preference alignment packaging • Multi-turn response evaluation, ranking, and cross-verification • Safety-aware annotation criteria enforcement for training data quality.

2024 - Present

Senior Data Annotation Engineer & Project Lead — SAM 2 Assisted Segmentation

SamaVideoSegmentation

Produced temporal semantic and instance segmentation labels for video data using AI-assisted SAM 2 workflows. Applied prompt-based interaction combined with inter-frame interpolation to accelerate pixel-level labeling. Maintained high QA pass rates while boosting segmentation efficiency for model iteration cycles. • AI prompt click workflows for SAM 2 • Inter-frame interpolation to propagate labels across video frames • Pixel-level semantic/instance segmentation annotation • QA monitoring and HITL calibration to sustain >99.5% pass rates.

2021 - 2024

Senior Data Annotation Engineer & Project Lead — CV Annotation Projects

OtherBounding Box

Managed oriented object detection annotation workflows for drone and fish-eye camera imagery using YOLO OBB and DOTA formats. Delivered high-precision oriented bounding boxes at production scale while sustaining strict quality assurance. Implemented pre-labeling and HITL processes to reduce manual effort without degrading label accuracy. • Rotated/oriented bounding box labeling in YOLO OBB and DOTA • Multi-object tracking dataset preparation (MOT/MOTS) • Large-batch ingestion and JSON/XML parsing for 10,000+ images per run • Python-driven cleaning/normalization and lossless format conversion for downstream training.

2021 - 2024

Education

Work History