Research Intern

About the Role:

As a Research Intern, you'll work on the computer vision and deep learning that power automated takeoff and estimation. The core problem is unsolved: a plan set is hundreds of pages drafted to dozens of CAD conventions, where the same primitive is a wall, a hatch pattern, or a leader line depending on context a model has to infer.

You'll own one scoped research problem end to end, paired with a senior researcher — framing it, running the experiments, and building a prototype evaluated on real data. It won't be a side project; it'll be something on the team's roadmap.

What You'll Do:

• Own one research problem end to end: framing, data, experiments, and a prototype that runs on real drawings
• Train and evaluate deep learning models for object detection, segmentation, or structured extraction on vector and raster construction drawings
• Explore self-supervised and representation learning on our corpus of unlabeled drawings
• Work with multimodal signals, combining visual layout, geometry, and text
• Design the evaluation for your problem, and present results to the research team

What We're Looking For:

• Currently pursuing a Master's, or PhD in Computer Science, AI/ML, Data Science, Mathematics, Electrical Engineering, or a related field
• Strong foundations in Python and at least one deep learning framework (PyTorch or TensorFlow)
• Coursework or project experience in computer vision, image processing, or machine learning
• Curiosity about applied research and a willingness to iterate quickly on ideas
• Comfortable working in a fast-paced, startup-like environment with close collaboration
• Clear written and verbal communication skills

Nice to have:

• Prior research experience, publications, or projects in computer vision or deep learning
• Familiarity with vector graphics and CAD formats (SVG, DXF, DWG, IFC)
• LLM and NLP experience like structured extraction, retrieval over long documents
• Experience with cloud environments (GCP, AWS, or Azure)

How We Work:

• Empirical over theoretical — we'd rather run the experiment than argue about it
• Truth-seeking, including about our own results. A negative result reported early is worth more than a positive one defended late
• Research is judged by whether it holds up in front of 1,200 contractors, not by whether it's clever
• Honest feedback, given directly

Compensation:

The stipend range for this role is 80-90$ per hour. The final offer depends on experience and level.