About
Machine Learning for Vision, Geospatial Intelligence, and 3D Data Products
I am a Machine Learning Scientist with a Ph.D. in Geomatics Engineering from the University of Calgary. My work combines deep learning, Bayesian inference, and scalable data engineering to solve problems in image understanding, LiDAR analytics, and geospatial modeling.
At Spatial Data Analytics Inc., I build production-oriented pipelines for image segmentation, object detection, and 3D point-cloud classification. My recent work spans TensorFlow and Keras segmentation models, Detectron2-based detection workflows, LiDAR feature engineering, and benchmarking across both classical and deep 3D models.
I also founded Flyover Map 3D, where I built a cloud application that transforms user routes into rendered 3D flyover videos using FastAPI, Docker, Cloud Run, Google Cloud Storage, Firestore, and large-scale geospatial preprocessing workflows.
Experience
Selected Roles
Senior Innovation Developer | Spatial Data Analytics Inc. | 2022 - Present
I design computer-vision and geospatial ML workflows for production and R&D. This includes image segmentation and object detection for utility and infrastructure imagery, preprocessing pipelines for tiling and normalization, feature extraction for LiDAR classification, and Dockerized training and inference workflows that integrate with engineering products.
Founder | Flyover Map 3D | 2025 - Present
I built a public web application that converts user routes into 3D flyover videos. The platform combines geospatial preprocessing, 3D mesh generation, cloud job orchestration, billing, and delivery workflows into a complete user-facing product.
Postdoctoral Researcher and Research Assistant | University of Calgary | 2017 - 2023
My academic work focused on Bayesian inversion, uncertainty estimation, anomaly-focused interpretation, and adaptive spatial modeling. I published peer-reviewed research on trans-dimensional inversion and hierarchical geophysical modeling while developing computational methods for large 3D scientific datasets.
A cloud-based geospatial product that transforms user routes into rendered 3D flyover videos through automated preprocessing, terrain modeling, rendering, and delivery.
Production segmentation workflows built with TensorFlow and Keras for utility and infrastructure imagery, using pretrained encoders, custom decoders, and task-specific losses.
Benchmarking and applied development for 3D point-cloud classification using PointNet++, PointConv, and classical baselines for LiDAR data.
Scalable preprocessing and inference workflows for imagery and LiDAR, covering tiling, normalization, feature engineering, Dockerized deployment, and cloud execution.
A research-driven extension of Voronoi partitioning for hierarchical spatial classification, designed to represent structure at multiple scales rather than as a single flat partition.
An adaptive geometry modeling approach in which the number and position of Voronoi cells are inferred from data, enabling flexible representation of unknown shapes and anomalies.
A data-driven approach for representing convex and concave boundaries through alpha-shape geometry, useful when the underlying structure is only partially known.
An educational Python automation project that explored DOM targeting, modal handling, scrolling strategies, and repeatable browser workflows using Selenium.
A Bayesian framework for subsurface interpretation from gravity and magnetic data, using adaptive partitioning and uncertainty-aware inference for 3D earth models.
Computer-vision work using Detectron2 for instance segmentation and object detection, with practical applications in infrastructure imagery and applied ML experimentation.
Motion estimation in video and satellite imagery using dense optical flow, with applications ranging from visual tracking to sea-surface current interpretation.
A high-performance LiDAR classification workflow built on engineered geometric features, Cython acceleration, and ensemble learning for point-cloud labeling.
Programming and Tools
- Python, C++, Cython, Julia, R, JavaScript
- FastAPI, Docker, Google Cloud, AWS, CUDA
- Git, Jupyter, LaTeX, parallel computing, MPI
ML and Computer Vision
- PyTorch, TensorFlow, Keras, Scikit-learn, Detectron2, SAM
- EfficientNetV2, DenseNet, MobileNetV2, Mask R-CNN, PointRend, Panoptic-DeepLab
- Semantic segmentation, object detection, anomaly detection, experiment design, benchmarking
Geospatial and 3D Data
- LiDAR point clouds, PointCONV workflows, feature engineering
- PDAL, GDAL/OGR, GeoPandas, Rasterio, PostGIS, Earth Engine
- Geospatial preprocessing, reprojection, mesh generation, scalable inference
The production frameworks I work with span TensorFlow/Keras segmentation pipelines, Detectron2-based instance segmentation and detection, and 3D point-cloud classification workflows ranging from feature-engineered random forests to PointNet++ and PointCONV benchmarking.
Awards
Scholarships and Recognition
- Alberta Graduate Excellence Scholarship, 2021
- Alberta Graduate Excellence Scholarship, 2019
- Helmut Moritz Graduate Scholarship, 2019
- External Awards: Miscellaneous, 2019
- 3rd Place at the Geomathon Competition, 2019
- Teaching Award, 2018
These awards reflect academic performance, research contribution, and teaching excellence during my graduate studies in Geomatics Engineering.