AI/ML Engineer – Architectural Drawing Understanding (US)

Responsibilities

• Develop and optimize computer vision models (classical + deep learning) for entity detection, segmentation, symbol recognition, and annotation extraction from architectural drawings.
• Apply classical CV techniques (e.g., edge detection, contour analysis, Hough transform, morphological operations) alongside deep learning models to solve vector and raster understanding tasks.
• Design and train deep learning models (e.g., CNNs, Mask R-CNN, U-Net, YOLO, DETR, Vision Transformers) for detection and segmentation of CAD drawing elements.
• Implement OCR pipelines for text and dimension extraction in drawings.
• Build robust data pipelines: preprocessing DWG files, rasterization/vectorization, augmentation, and dataset creation for supervised training.
• Benchmark, evaluate, and continuously improve model accuracy, robustness, and efficiency.
• Collaborate with cross-functional teams to integrate vision models into design automation and CAD/BIM workflows.

Qualifications

• Education & Background: Bachelor’s, Master’s, or PhD in Computer Science, Artificial Intelligence, Computer Vision, or related fields. Strong foundation in mathematics, geometry, and image processing.
• Computer Vision Expertise (Priority):
  • 3+ years of hands-on experience building CV pipelines and production-ready ML models.
  • Proven track record with classical CV algorithms (OpenCV, scikit-image): contour/edge detection, shape matching, geometric transformations, Hough transform, morphological filtering.
  • Strong experience training and deploying deep learning CV models: CNNs, U-Net, Mask R-CNN, Faster R-CNN, YOLO, DETR, Vision Transformers, SAM, etc.
  • Experience with OCR (e.g., Tesseract, deep-learning-based text recognition).
  • Practical experience in combining classical CV with deep learning for hybrid solutions.
• Technical Skills:
  • Proficiency in Python and ML frameworks (PyTorch, TensorFlow).
  • Strong engineering practices: Git, CI/CD, testing, Docker, and scalable inference deployment.
  • Familiarity with vector graphics, CAD data formats (DWG/DXF), and computational geometry is a plus, but not mandatory.
• Preferred Skills:
  • Knowledge of geometric deep learning or graph-based approaches for structured vector data.
  • Experience with annotation tools, dataset creation, and augmentation for CV tasks.
  • Familiarity with AEC (Architecture, Engineering, Construction) workflows is an advantage.