Mi Wei

Key Skills

Software

Appen

CVAT

LabelImg

Label Studio

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Image

Video

Top Task Types

Bounding Box

Classification

Land Cover Classification

Mapping

Polygon

Freelancer Overview

A data labeling expert with a strong focus and extensive experience in autonomous vehicle data labeling and ASR speech-to-text transcription. Proficient in image labeling, LiDAR point cloud labeling, and sensor fusion labeling, as well as Mandarin Chinese speech-to-text transcription. Familiar with various labeling tools and quality control processes. Skilled at collaborating with efficient teams to provide high-quality training data for machine learning models.

Labeling Experience

2D/3D bounding boxes, instance segmentation

Internal Proprietary Tooling3D SensorBounding BoxPoint Key Point

Annotated 30,000+ frames for camera-based object detection (2D/3D bounding boxes, instance segmentation). Processed 500+ LiDAR point cloud scenes with 3D cuboids, track IDs, and occlusion flags; achieved 98.7% IoU consistency.

Annotated 30,000+ frames for camera-based object detection (2D/3D bounding boxes, instance segmentation). Processed 500+ LiDAR point cloud scenes with 3D cuboids, track IDs, and occlusion flags; achieved 98.7% IoU consistency.

2024

ASR Transcription & Calibration Dataset for Hospital Medical Dialogues

Internal Proprietary ToolingAudioTranslation LocalizationAudio Recording

In collaboration with a leading AI healthcare unicorn, we delivered over 100,000 real-world medical recordings (totaling 250 hours), all in Mandarin. Core tasks included: word-for-word transcription (WER < 5%), phoneme-level temporal alignment, standardized annotation of medical terminology (drug names, symptom descriptions), noise filtering, speaker separation, and multi-turn dialogue intent slot filling.

In collaboration with a leading AI healthcare unicorn, we delivered over 100,000 real-world medical recordings (totaling 250 hours), all in Mandarin. Core tasks included: word-for-word transcription (WER < 5%), phoneme-level temporal alignment, standardized annotation of medical terminology (drug names, symptom descriptions), noise filtering, speaker separation, and multi-turn dialogue intent slot filling.

2024 - 2024

drivable area data

Internal Proprietary Tooling3D SensorPolygonPolyline

Labeled 40,000+ images for lane detection, traffic sign classification, and pedestrian intent prediction. Created polygon-level semantic segmentation masks for drivable area and lane markings.

Labeled 40,000+ images for lane detection, traffic sign classification, and pedestrian intent prediction. Created polygon-level semantic segmentation masks for drivable area and lane markings.

2023 - 2024

multi-sensor data (camera + LiDAR + radar)

Appen3D SensorBounding BoxSegmentation

Annotated over 25,000 frames of multi-sensor data (camera + LiDAR + radar) for training perception models in autonomous vehicles. Performed 3D bounding box labeling on LiDAR point clouds and 2D polygon annotation on camera images for object detection tasks. Conducted sensor fusion annotation to align objects across multiple modalities, ensuring temporal and spatial consistency. Collaborated with ML engineers to refine annotation guidelines and improve label accuracy and consistency. Participated in quality assurance workflows, achieving >98% annotation accuracy in internal

Annotated over 25,000 frames of multi-sensor data (camera + LiDAR + radar) for training perception models in autonomous vehicles. Performed 3D bounding box labeling on LiDAR point clouds and 2D polygon annotation on camera images for object detection tasks. Conducted sensor fusion annotation to align objects across multiple modalities, ensuring temporal and spatial consistency. Collaborated with ML engineers to refine annotation guidelines and improve label accuracy and consistency. Participated in quality assurance workflows, achieving >98% annotation accuracy in internal

2021 - 2023

Education

Work History