SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 90,000 - 125,000 / annual
Bobyard is a Series A AI startup building visual intelligence for construction. The company's models are trained on millions of construction drawings to help contractors estimate and bid faster. You'll own the data labeling operation end-to-end as a first-in-function hire, building the playbook from scratch.
You'll manage the annotator team—recruiting, onboarding, training, and maintaining quality and throughput standards. You'll own labeling quality by reviewing annotations, catching systematic errors before they reach models, and refining guidelines based on findings. A significant part of the role involves cleaning up existing datasets: fixing inconsistent labels, missing metadata, duplicates, and other issues that hurt model performance.
You'll source new construction drawings to expand coverage of formats, classes, and edge cases the models currently miss. You'll translate ML engineer requests into shipped datasets by scoping asks, running projects, and delivering clean data on schedule. Working directly with ML engineers, you'll understand where models fail and build the data needed to fix those failures.
Beyond people management, you'll build tooling and workflows that make labeling faster and more reliable. Success means a labeling operation that's fast, accurate, and trusted enough that ML engineers stop double-checking your team's work.
You should have direct experience managing a labeling, annotation, or data-quality team. You're extremely detail-oriented, noticing data issues before others point them out. You have strong operational instincts to juggle multiple datasets, annotators, and priorities without losing details. You're technical enough to work with ML engineers—understanding false positives, false negatives, class imbalance, and train/test splits—and can set up your own tools to speed up labeling. You're resourceful when sourcing new data types and take high ownership to ensure datasets are actually good.
Nice-to-have skills include familiarity with labeling platforms (Labelbox, CVAT, Supervisely), basic SQL or Python for data querying and cleaning, and background in construction, CAD, or other visually complex technical domains.
