> Markdown version of [/jobs/ext/4210325-data-annotation-specialist-computer-vision](https://www.wearedevelopers.com/jobs/ext/4210325-data-annotation-specialist-computer-vision). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Data Annotation Specialist - Computer Vision

> This ad is more than 30 days old and may be closed.

- **Company:** BOBYARD INC.
- **Location:** San Francisco, CA, United States
- **Salary:** $41,600 - $72,800
- **Contract:** Temporary contract
- **Skills:** Computer-Aided Design, Computer Vision, Computer Literacy, Object Detection
- **Published:** October 8, 2026
- **Apply:** https://www.indeed.com/viewjob?jk=9c3e998527fde73a

## About the Role

* You catch what's wrong, inconsistent, or missing before anyone points it out
 * You follow written instructions precisely, and use good judgment when they run out
 * You can do focused, repetitive work for hours without your accuracy slipping
 * You learn new tools fast and work well on your own, remotely
 * Clear written English for flagging issues and asking questions

Nice to have

 * Prior image annotation work in Labelbox, CVAT, or Supervisely
 * Basic computer vision concepts: object detection, segmentation
 * You can read construction, CAD, or architectural drawings

## Description

You'll label the construction drawings our computer vision models learn from, and work directly with the ML engineers who use your data. We're adding new trades, so our models need more labeled drawings than ever. The work is repetitive and exacting, and the guidelines won't cover every case. Success is labels so consistent that engineers use them without a second look.

What you'll do

 * Label construction drawings with bounding boxes, polygons, and segmentation masks that meet our accuracy bar
 * Apply guidelines consistently, so labels match from the first drawing to the last
 * Flag ambiguous and edge cases early, and turn them into sharper guidelines
 * Clean existing datasets: fix inconsistent labels, missing metadata, and duplicates
 * Hit weekly quality and throughput targets
 * Tell ML engineers what you see in the data, so they fix failures faster