About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. OpenTrain AI is the hiring and contracting organization for this role, connecting contributors with practical projects that help build modern artificial intelligence.

• Create an OpenTrain account for free.
• Apply through OpenTrain for remote AI training and data-labeling work.
• Build experience by contributing to datasets used to improve AI systems.

About AI Training and Data Labeling

AI systems learn from examples prepared and reviewed by people. In this project, you will help create structured visual data by identifying engineering symbols and mapping the relationships between them.

AI training and data-labeling work is a growing part of the technology industry. Projects are often remote and flexible, giving contributors opportunities to apply specialized knowledge while helping shape how AI understands real-world information.

• Work with image-based data and detailed annotation guidelines.
• Contribute labels that support object-detection and graph-based machine-learning datasets.
• Use careful human review to make AI training data more accurate and consistent.

The Role

We are seeking an intermediate engineering diagram graph annotator to label approximately 60 sheets. Each diagram must be represented as a graph, with every relevant symbol enclosed in a classified bounding box and every connection between symbols labeled as an edge.

The project begins with a small paid one-sheet pilot before the full batch. The total fixed project price is $499. You will work as a contractor.

• Project type: Image annotation and graph relationship labeling
• Estimated volume: Approximately 60 sheets
• Compensation: $499 fixed price
• Engagement: Contractor
• Work arrangement: Worldwide and remote

What You'll Annotate

Annotations must be drawn at full resolution and aligned to the supplied images. Symbol boxes should tightly enclose the symbol glyph rather than the connecting line, while edges should represent connectivity between nodes.

The final annotation tool and export format are still being determined. Expected formats include Label Studio JSON, CVAT XML, or COCO, and downstream conversion will be handled separately.

• Create bounding boxes for valve, pump, instrumentation, general, tank, arrow, and inlet/outlet symbols.
• Label connector routing or helper nodes, including line bends and junctions.
• Label crossing nodes where lines intersect.
• Draw edges between connected nodes.
• Classify each edge as solid or non-solid.
• Keep labels consistent with the written guideline across the complete batch.

Required Experience and Skills

You should be comfortable reading technical or engineering diagrams and distinguishing symbols from the lines that connect them. Experience with an annotation tool that supports both bounding boxes and region-to-region relations is required.

The work involves large, dense images, so careful zooming, precise placement, and repeatable quality-control habits are important. Prior object-detection or graph and relationship dataset annotation experience is helpful but not required.

• Hands-on experience with Label Studio, CVAT, or an equivalent tool supporting bounding boxes and relations.
• Ability to label connections between regions, not only individual objects.
• Comfort reading technical or engineering diagrams.
• Strong attention to detail when working at full image resolution.
• Ability to follow a fixed class taxonomy and written annotation guideline.
• Consistent, reliable work across a multi-sheet batch.

Evaluation and Quality Expectations

Candidates will be assessed on tool proficiency, diagram comprehension, classification consistency, bounding-box precision, and connectivity labeling. The evaluation also considers realistic throughput, self-review practices, and responsiveness to quality-control feedback.

For ambiguous symbols, the expected approach is to flag the case for review or request clarification rather than guess. If quality control identifies loose boxes or missing edges, annotators should apply the feedback systematically and prevent the same issue from recurring.

• Explain which annotation tools you have used and where you created both boxes and region relations.
• Describe how you separate a tight symbol box from the connecting edge.
• Explain how you handle symbols that do not clearly match the provided legend.
• Describe how you maintain pixel-accurate boxes on large, dense images.
• Share an example of labeling relationships or connections between objects.
• Give a realistic estimate of how many densely labeled sheets you can complete per week.
• Describe your self-QC process, including reviewing connectivity and avoiding fatigue errors.
• Be prepared to respond constructively to QC corrections.

How the Project Works

The engagement starts with a small paid one-sheet pilot to confirm annotation quality and guideline alignment. Successful pilot work can lead into the broader batch of approximately 60 sheets.

You will provide clean, consistent labels in the agreed format and coordinate your work with the written guidelines. OpenTrain AI will handle downstream format conversion if needed.

• Apply through OpenTrain.
• Complete the paid one-sheet pilot.
• Annotate the full batch if the pilot meets the acceptance criteria.
• Use the agreed annotation workflow and coordinate on any tool or format updates.
• Submit consistent graph annotations for review.