Senior/Staff 3D Reconstruction Engineer, Spatial AI - DroneDeploy

Senior/Staff 3D Reconstruction Engineer, Spatial AI - DroneDeploy Over the last ten years, DroneDeploy has built one of the most hardened photogrammetry pipelines in the world. It turns drone, ground, and 360° captures from over 3 million jobsites into maps, models, and measurements that construction teams rely on every day. We are now building its successor: a reconstruction stack that moves reality capture from hours to minutes and delivers substantially higher-quality 3D, built on learned methods like Gaussian splatting, feed-forward 3D foundation models (VGGT, MASt3R, MapAnything and their successors), learned matching and depth, and streaming reconstruction. We are looking for a Senior or Staff engineer with deep specialist expertise in modern 3D reconstruction and spatial AI to design and build the core building blocks of that next-generation pipeline. You will be the team's go-to expert in your area, shape the 3D reconstruction roadmap, and turn state-of-the-art research into reliable product features used on construction sites around the world. This is a hands-on individual contributor role with real technical leadership and no people-management expectation. Your impact comes from what you build, the technical direction you set, and how you raise the bar for the engineers around you. Within your first 3-6 months, we expect a learned component you built to be running in production, at scale, on customer data. What You'll Do • Build core reconstruction building blocks. Design, train, and ship learned 3D components that form the foundation of our pipeline: pose and geometry estimation, dense reconstruction, scene representation, and rendering. • Bring next-generation 3D methods to production. Lead evaluation and integration of splatting, neural rendering, and 3D foundation models. Decide where they replace, augment, or initialize classical SfM and MVS. • Push toward real-time reality capture. Develop fast, incremental, and streaming reconstruction that shortens the time from capture to usable 3D across drone, ground, and 360° capture. • Own model architecture and training. Design architectures, losses, and training pipelines. Curate datasets from our capture corpus. Fine-tune and adapt foundation models to our domain. • Shape the technical roadmap. Partner with ML leadership to identify which research bets to make, size their product impact, and turn them into concrete milestones. • Set the bar for evaluation. Define benchmarks, metrics, and evaluation datasets for geometric accuracy, visual quality, robustness, and runtime. New methods win on evidence, not demos. • Take research to production. Work with the engineers who own pipeline infrastructure to deploy models at scale on GPU cloud infrastructure, with attention to cost, latency, and reliability. • Lead technically by example. Write design docs, give thoughtful code and architecture reviews, run reading groups, and help engineers level up in 3D ML. • Stay at the frontier. Track CVPR, ICCV, ECCV, SIGGRAPH, and NeurIPS, prototype promising work quickly, and share what you learn. • Use AI tooling every day. AI coding agents (e.g., Cursor, Claude Code) are a core part of how we work. Help define how the team uses them for research and engineering, and hold agent-written code to the same standard as human-written code. What You Bring • Deep, demonstrated expertise in at least one of the following, with working knowledge of the others: • Gaussian splatting, NeRF-family methods, or differentiable and neural rendering • Feed-forward 3D foundation models, learned SfM, or pose and depth regression • Streaming or real-time learned reconstruction (e.g., CUT3R, MASt3R-SLAM, splatting-based SLAM), beyond classical SLAM • Learned feature matching, learned multi-view stereo, or monocular depth estimation • Strong grounding in multi-view geometry: camera models, bundle adjustment, SfM, and MVS. You know when classical methods beat learned ones, and when to combine them. • Hands-on experience designing, training, and debugging 3D deep learning models in PyTorch. • A track record of taking research ideas into working systems: production features, widely used open-source code, or strong published results. • 5+ years of industry or applied research experience in 3D computer vision or 3D ML (a PhD in a related area counts). Master's or PhD in Computer Science, Electrical Engineering, Robotics, or a related field, or equivalent practical experience. Staff level additionally brings 8+ years, a record of setting technical direction across a team or organization, and ownership of multi-quarter technical bets from research question to shipped product. Nice to Have • Large-scale outdoor, aerial, or geospatial reconstruction, including georeferencing and accuracy against ground truth. • 360° cameras, including spherical and equirectangular camera models. • Modifying, fine-tuning, or retraining large 3D foundation models for new domains, data, or camera types. • Mesh extraction from splats or radiance fields, surface reconstruction, or texturing. • Large-scale multi-GPU training infrastructure. • C++ or CUDA for performance-critical code. • Publications at top venues (CVPR, ICCV, ECCV, SIGGRAPH, NeurIPS, 3DV) or notable open-source contributions (e.g., gsplat, COLMAP). • Real-time or on-device 3D (ARKit, ARCore, mobile GPUs). • Shipping GPU workloads on cloud platforms (AWS, GCP, Azure) with Docker and Kubernetes. Additional Information Base Pay Range: 140,960.00 - 193,820.00 USD Annual This role may also be eligible for Equity Compensation and/or Bonus Incentive Compensation. Procore is committed to offering competitive, fair, and commensurate compensation. Actual compensation will be based on a candidate's job-related skills, experience, education or training, and location.