Computer Vision Research Engineer
At a glance
Mid-level AI Research role at Bobyard. San Francisco · full-time · $160,000–$230,000 base.
aimlroles summary, based on the employer's posting.
What you'll do
- Develop detection, segmentation, OCR and drawing-reasoning models where errors can have financial consequences
- Run model pipelines from incoming data and labeling through training, evaluation and fast inference
- Raise accuracy, resilience and speed across varied drawing formats, then investigate customer-data failures
- Work with product and fullstack teams to put models into customer-facing features
- Review new research and quickly apply useful methods in shipped work
What you bring
- You have trained and shipped computer vision or deep learning models, such as detection, segmentation or document AI
- You have put machine learning into production where dependable performance matters
- You can work with imperfect inputs, including scanned or handwritten material
Who this fits
This mid-level, hands-on role suits an engineer who takes ownership of model performance and enjoys shaping incomplete requirements. You’ll work onsite in San Francisco, in a weekly shipping cycle where priorities can shift as customer feedback comes in.
From the employer
Position Overview
Join the team building computer vision applications for the construction trades.
Contractors spend dozens of hours per project doing takeoffs by hand. We're automating that — with models that need to perform reliably on messy, real-world drawings.
You'll have a lot of autonomy to design, execute, and iterate. We move quickly, and you'll build deep expertise in the models you ship. We're looking for strong CV/ML engineers with good product judgment who are comfortable both designing solutions and doing the hands-on work to ship them.
Responsibilities
Build strong CV models: detection, segmentation, OCR, reasoning over construction drawings. Accuracy matters — mistakes can cost a customer real money.
Own end-to-end pipelines: data ingestion, labeling, training, eval, and low-latency inference.
Improve performance across accuracy, robustness, and latency as we handle a wide range of drawing formats.
Ship to production — work closely with fullstack and product to get models into features customers use daily.
Dig into real customer data to find and fix failure modes.
Stay current with the field and apply what's useful quickly.
Desired Attributes
High ownership — you take responsibility for model performance without needing to be told what to fix
Strong CV + deep learning background — you've trained and shipped models (segmentation, detection, document AI, or vision-language)
Experience putting ML into production with real reliability requirements
Comfortable with messy, real-world data — scanned, handwritten, imperfect
Practical mindset — you read the research, but your focus is on shipping working code
Comfortable in a fast-moving, early-stage environment
A note on pace
We ship weekly, and priorities move as we learn from customers. Requirements are often incomplete when work starts, and you'll be expected to shape them. If that pace excites you, you'll fit in well.
Benefits
Competitive salary plus equity options
Health, dental, and vision coverage
Catered lunch in the office
$70 a month toward your commute
Team happy hours and events
Build a category-defining company with a small, driven team
Come build with us!