Francisco Castellano

Key Skills

Software

Remotasks

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Action Recognition

Bounding Box

Cuboid

Polygon

Segmentation

Freelancer Overview

Bilingual specialist with 5+ years of experience in AI training data for autonomous vehicles. Expert in LiDAR annotation and image segmentation using Scale AI/Remotasks, with a proven track record of delivering high-precision datasets for computer vision. My unique value comes from a dual skill set: deep technical labeling expertise combined with Zendesk-based support and workflow management experience, gained through an internal promotion. This allows me to not only execute complex annotation tasks but also contribute to team efficiency and project success from an operational perspective.

Labeling Experience

Lidar

Scale AI3D SensorCuboid

Spearheaded the annotation of 3D LiDAR point cloud sequences for an autonomous vehicle perception system, focusing on complex urban driving scenarios. Executed instance and semantic segmentation on point clouds to identify and classify objects such as vehicles, pedestrians, cyclists, and static infrastructure with high geometric precision. Implemented object tracking across frames to provide consistent trajectory data, crucial for path prediction algorithms. Ensured data quality and consistency by adhering to complex project specifications, directly contributing to a training dataset that achieved a >98% validity score in client audits. Collaborated with QA analysts to refine labeling guidelines, improving the efficiency and accuracy of the annotation pipeline for 3D sensor data.

Spearheaded the annotation of 3D LiDAR point cloud sequences for an autonomous vehicle perception system, focusing on complex urban driving scenarios. Executed instance and semantic segmentation on point clouds to identify and classify objects such as vehicles, pedestrians, cyclists, and static infrastructure with high geometric precision. Implemented object tracking across frames to provide consistent trajectory data, crucial for path prediction algorithms. Ensured data quality and consistency by adhering to complex project specifications, directly contributing to a training dataset that achieved a >98% validity score in client audits. Collaborated with QA analysts to refine labeling guidelines, improving the efficiency and accuracy of the annotation pipeline for 3D sensor data.

2017 - 2020

Basic segmentation

Scale AIImageSegmentation

Scope: Create a pixel-perfect labeled dataset to train an AV's scene-understanding AI across diverse urban environments. Tasks: Executed manual, pixel-level semantic segmentation on images, labeling 10+ object classes (road, vehicles, pedestrians, etc.) using the Scale AI - Remotasks platform. Size: Annotated 50,000+ images as part of a team over a 6-month project cycle. Quality: Adhered to a strict >99% accuracy standard, passed multi-stage reviews (peer + QA), and maintained high consistency via regular team calibration sessions.

Scope: Create a pixel-perfect labeled dataset to train an AV's scene-understanding AI across diverse urban environments. Tasks: Executed manual, pixel-level semantic segmentation on images, labeling 10+ object classes (road, vehicles, pedestrians, etc.) using the Scale AI - Remotasks platform. Size: Annotated 50,000+ images as part of a team over a 6-month project cycle. Quality: Adhered to a strict >99% accuracy standard, passed multi-stage reviews (peer + QA), and maintained high consistency via regular team calibration sessions.

2018 - 2018

Education

Work History