AI/ML Engineer – LiDAR Asset Extraction & Multi-Epoch Change Detection

AI/ML Engineer – LiDAR Asset Extraction & Multi-Epoch Change Detection We are seeking an AI/ML Engineer with demonstrated prior experience working with LiDAR point-cloud datasets to develop AI capabilities for extracting and detecting changes in transportation infrastructure assets. The engineer will work with multi-epoch LAS/LAZ point clouds collected over the same transportation corridors at different points in time. Candidates must have hands-on experience with LiDAR processing and AI/ML frameworks, preferably including Python, PyTorch/TensorFlow, PDAL, Open3D, PointNet/PointNet++, KPConv, sparse 3D CNNs, or related point-cloud technologies. The engineer will implement four core capabilities: (1) DEM/DSM differencing to detect elevation and surface changes between LiDAR epochs; (2) voxel-based 3D change detection to identify added, removed, and modified features; (3) object-based change detection to extract and compare transportation assets such as roads, bridges, guardrails, signs, poles, barriers, and other roadside infrastructure; and (4) AI/deep-learning point-cloud segmentation to automatically classify transportation assets from LiDAR. The solution must address practical issues including coordinate normalization, registration, varying point densities, noise, occlusion, classification uncertainty, and differences between acquisition years. The solution will be developed and deployed on Google Cloud Platform (GCP) and should support scalable processing of large LiDAR datasets. The engineer will design cloud-based pipelines for ingesting LAS/LAZ data from Cloud Storage, preprocessing and tiling point clouds, generating DEM/DSM and voxel representations, training and executing AI models using appropriate GCP compute/GPU resources, and storing extracted assets and change-detection results for consumption by downstream SaaS applications and APIs. The implementation should be modular, reproducible, containerized, and capable of scaling from small proof-of-concept datasets to transportation-corridor and DOT-scale datasets. **Expected outcome:** a working GCP-based AI pipeline that accepts **two or more LiDAR datasets of the same geographic location acquired in different years** and produces both transportation asset inventories and measurable changes between epochs. Outputs should identify assets as **New, Removed, Modified, or Unchanged**, provide location/geometry and confidence scores, generate DEM/DSM difference and 3D change layers, and expose results in formats suitable for GIS visualization and downstream transportation asset-management systems. The successful engineer should be able to demonstrate the solution using representative transportation LiDAR datasets and provide documented model performance, processing workflow, source code, deployment instructions, and reproducible benchmark results.