Senior Machine Learning Engineer

Spexi•Vancouver, BC

1d•Remote

About The Position

Spexi is a drone technology company on a mission to make ultra-high-resolution geospatial imagery more accessible than ever before, empowering humanity to make better decisions about the physical world. We’re building an exciting new two-sided marketplace called the Spexi Network, powered by drones and blockchain technology. It's the world’s first Fly-to-Earn platform that enables drone pilots to earn rewards for flying and collecting aerial imagery. It also enables organizations of all sizes to quickly and easily access high-resolution aerial imagery and valuable derivative data, powering remote monitoring of buildings, infrastructure, natural resources, and more. Our goal is to guide their decision-making and help them better plan and react, without needing to own drones or hire pilots. We’re looking for a Senior Machine Learning Engineer to lead the development of experimental models, algorithms and prototype systems that push the boundaries of what’s possible with geospatial imagery analytics. Your work will bridge early-stage research and production—delivering high-quality, well-structured code that serves as a foundation for the next generation of Spexi’s geospatial intelligence products.

Requirements

An M.S. and 5+ years of work experience in Computer Science, Computer Vision, Machine Learning, Remote Sensing, or a related quantitative field.
Proven experience in research-to-production translation: acting as the bridge between pure academia and commercial engineering - demonstrating the ability to read a newly published research paper, replicate its findings, rapidly prototype, and distill it into a production-ready feature.
At least 3 years of applied ML and computer vision experience transitioning models from research to production, ideally involving geospatial, aerial, or satellite imagery.
Deep, contemporary expertise in predictive AI for imagery, including classification, object detection, and segmentation, with a strong technical perspective on the efficacy of modern feature-extraction methods.
Working knowledge of the SAM2 or similar algorithms—including fine-tuning, prompt design, and distillation—and a clear understanding of its strengths and failure modes within the context of aerial datasets.
Hands-on experience developing Generative AI capabilities, such as multimodal RAG, vision-language models, and diffusion-based pipelines, with the ability to ship complete systems end-to-end.
Strong technical judgment regarding the selection of predictive vs. generative approaches, considering critical factors like cost, latency, and evaluability.
Production experience operating ML pipelines on AWS, specifically utilizing SageMaker for training and hosting, alongside broader AWS data orchestration services.
Experience managing large, complex imagery datasets in cloud environments while optimizing models for real-world performance, throughput, and cost-efficiency.

Nice To Haves

A Ph.D. and 7+ years of work experience blending artificial intelligence with the physical sciences (e.g., Photogrammetry, Physics)
Deep expertise in geospatial and remote sensing workflows, including hands-on experience with georeferenced imagery, coordinate reference systems, projections, and industry-standard tools like GDAL, PostGIS, and rasterio.
Proven ability to adapt or pretrain geospatial foundation models to specialized remote-sensing tasks.
Specialized experience in 3D scene understanding, leveraging NeRFs, Gaussian Splatting, and point cloud segmentation to ensure multi-view consistency across 2D and 3D ML outputs.
Architectural experience designing multimodal RAG systems that integrate imagery, vector, and time-series data, with a focus on rigorous retrieval and generation benchmarks.
Background in fast-paced startup environments, with a demonstrated capability to translate experimental research into production-quality geospatial intelligence systems.

Responsibilities

Design, train, and productionize models for aerial-image classification, object detection, and 2D/3D segmentation across diverse geographies, sensors, resolutions, and seasons.
Spearhead the integration and optimization of state-of-the-art foundation models for aerial segmentation, adapting SAM2-like capabilities and subsequent architectures through prompt engineering, fine-tuning, and distillation
Engineer change detection and structure-change models capable of distinguishing real-world physical changes from acquisition noise and seasonal lighting variations.
Develop predictive models for trend forecasting, integrating time-series methods with spatial context to monitor vegetation growth, construction, and asset degradation.
Build Generative AI capabilities, including multimodal models and natural-language query systems that ground language in georeferenced pixels and semantic layers.
Design and operate scalable ML pipelines on AWS, leveraging SageMaker, S3, and Step Functions to move from research prototypes to production endpoints.
Track the frontier of research—including NeRFs, Gaussian Splatting, and diffusion models—translating relevant breakthroughs into shipped product capabilities.
Collaborate with photogrammetry and platform teams to ensure ML outputs maintain geospatial accuracy and align to coordinate reference systems.
Establish rigorous evaluation benchmarks and metrics to validate model performance under real-world production conditions.