Computer Vision Engineer

About this role

CivilGrid seeks a Computer Vision Engineer to build CV/ML solutions for infrastructure planning software used by major utilities and engineering firms. You'll develop classical and deep learning-based vision systems to extract insights from messy real-world imagery and geospatial data, directly impacting product scalability in a Series A startup.

What you'll do

• Identify high-impact opportunities for computer vision and ML solutions within CivilGrid's data and workflows
• Implement classical CV techniques (feature matching, geometric transformations, image registration) for internal and customer-facing features
• Prepare datasets and train, fine-tune, and validate vision-based models including CNNs and transformers
• Develop hybrid solutions combining Vision Language Models, deep learning, and classical CV approaches
• Ship production-ready CV pipelines with built-in monitoring, evaluation harnesses, and quality metrics
• Collaborate across small team to solve problems affecting utilities, municipalities, and engineering firms

What they're looking for

• Python and computer vision libraries (OpenCV, NumPy, PyTorch or equivalents)
• Classical computer vision (feature extraction, geometric transformations, RANSAC, image registration)
• Deep learning and CV models (CNNs, transformers, embeddings)
• Dataset preparation and model training/fine-tuning/evaluation
• Vision Language Model (VLM) integration
• Geospatial tooling (GDAL, rasterio, coordinate reference systems)
• MLOps practices
• Real-world messy imagery handling (scanned documents, OCR, low-quality rasters)

Benefits

• Base salary $150,000–$215,000
• Meaningful early-stage equity
• Comprehensive benefits package
• Hybrid work (2+ days/week Boston office)
• Direct ownership and real impact from day one
• Proximity to tier-1 customers (PG&E, major utilities, engineering firms)

Likely interview questions

• Walk us through a computer vision project where you combined classical and deep learning approaches—what was the problem and why did you need both?
• Describe your experience handling messy or low-quality real-world imagery. What preprocessing or robustness techniques did you use?