Junior Lead ML Engineer - Computer Vision

You’ll work alongside experienced engineers to improve our AI-enabled plan ingestion pipeline—from raw PDFs to clean, dependable outputs that power takeoff, design, and downstream automation. You’ll begin by contributing to well-scoped projects and supporting senior engineers, with the opportunity to take on greater ownership as you grow.

Support the development and improvement of machine learning systems for object detection, segmentation, document understanding, and information extraction from building plans

Train, evaluate, and debug computer vision models using real-world construction data

Help build and maintain datasets, labeling workflows, preprocessing pipelines, and evaluation tools

Contribute to experiments involving computer vision, document understanding, and related machine learning techniques

Assist with integrating models into production applications and APIs

Write clean, testable, and maintainable Python code

Investigate model failures and help identify opportunities to improve accuracy and reliability

Collaborate with senior engineers to understand technical requirements and turn them into working solutions

Document experiments, results, decisions, and lessons learned

Learn and apply engineering practices for testing, deployment, observability, and maintainability

1–2 years of professional, internship, research, or equivalent project experience in machine learning, computer vision, or a closely related area

A degree in computer science, engineering, mathematics, data science, or a related technical field, or equivalent practical experience

Strong Python fundamentals and experience using PyTorch or a similar deep learning framework

Familiarity with computer vision tasks such as object detection, image segmentation, classification, or OCR

Basic understanding of image processing concepts and tools such as OpenCV

Familiarity with Linux and Git

Experience working with data, training models, evaluating results, and debugging failures

Willingness to ask questions, receive feedback, and learn from more experienced engineers

Clear communication skills and comfort working with a remote and cross-cultural team

We do not expect junior candidates to have experience with every technology listed in this description. Strong fundamentals, curiosity, and evidence that you can learn quickly matter more than checking every box.

You have strong technical fundamentals and are excited to apply them to real-world problems

You enjoy experimenting, debugging, and understanding why a model succeeds or fails

You take responsibility for your work while knowing when to ask for help

You care about writing clear, reliable code—not just producing promising model results

You are curious, motivated to improve, and comfortable working on problems without obvious solutions

You communicate clearly about your progress, questions, and blockers

Academic, internship, or personal project experience involving document understanding, OCR, or technical drawings

Experience with detection or segmentation frameworks

Familiarity with FastAPI, Docker, ClearML, or MLflow

Interest in multi-modal models, language models, NLP, or retrieval-augmented generation

Exposure to model deployment, inference optimization, or data-labeling workflows

A portfolio, GitHub repository, research project, or other examples of technical work

Screening call Online skills assessment 30-minute conversation with our CPO Technical interview with our CTO and engineering team

Benchmark (formerly BotBuilt, Y Combinator W21) is building the next generation of technology behind American homebuilding. We’re starting with one of the hardest parts of the stack: reliably turning messy, real-world construction plans into structured, usable data.