About Citylogix

Citylogix is a data and analytics company transforming how transportation infrastructure is managed. We build intelligent GIS and asset management solutions that help municipalities and engineering partners make smarter, evidence-based decisions. Combining startup agility with deep technical expertise, our team works at the intersection of geospatial technology and civil infrastructure to solve real‑world challenges and build smarter, more resilient communities.

About The Role

Reporting to the Chief Technology Officer (CTO), the Senior Machine Learning Engineer is responsible for researching, developing, and deploying the computer‑vision models at the core of our platform — the segmentation systems that detect infrastructure defects and map road assets from large‑scale imagery. This is a hands‑on technical role focused on transformer‑based segmentation. You will own models from research through production: improving accuracy and robustness, generalizing models across different camera systems, consolidating task‑specific models into unified architectures, and partnering with our DevOps and Engineering teams to bring them reliably into production at scale. What You’ll Do

Research, design, train, and evaluate deep‑learning models for semantic, instance, and panoptic segmentation, applied to infrastructure‑defect detection and road‑asset/scene segmentation Fine‑tune and adapt transformer‑based segmentation architectures (e.g., Mask2Former, OneFormer, and DETR‑family models, including custom in‑house variants) Improve domain generalization so models perform reliably across heterogeneous camera systems and imaging conditions, both individually and in combination Consolidate multiple task‑or defect‑specific checkpoints into unified,



multi‑task models to simplify the production footprint Partner with the DevOps team to train, deploy, and monitor models on AWS (including Amazon SageMaker), and to optimize inference performance and cost Build and maintain reproducible training pipelines, evaluation benchmarks, and dataset/annotation workflows Diagnose model failures and data‑quality issues, and drive measurable improvements in accuracy, robustness, and efficiency Track emerging architectures and techniques, and recommend where they create real value for the platform Maintain clear technical documentation, experiment records, and model‑versioning standards Qualifications

Minimum of 5 years of applied machine‑learning / computer‑vision experience, or a graduate degree (MSc/PhD) in a relevant field plus 3+ years of applied experience Strong, hands‑on proficiency with PyTorch Demonstrated experience with deep‑learning segmentation (semantic, instance, and/or panoptic) on real‑world image data Practical experience with transformer‑based vision architectures (Mask2Former, OneFormer, DETR, or similar) Solid command of fine‑tuning, transfer learning, and domain‑adaptation/generalization techniques Experience with the Hugging Face ecosystem and/or detectron2 Comfort training and deploying models in a cloud environment (AWS preferred; experience with Amazon SageMaker a strong asset) Solid Python engineering skills and good software practices (version control, testing, reproducibility)



Excellent communication and collaboration skills, and the ability to manage priorities in a fast‑paced environment Nice to Have

Experience working with imagery from industrial, line‑scan, or panoramic camera systems Background in geospatial / GIS, remote sensing, or infrastructure/civil applications MLOps

experience:
model lifecycle management, monitoring, and CI/CD for ML Model‑optimization experience (quantization, distillation, ONNX/TensorRT, or similar) for efficient inference Experience unifying multi‑task or multi‑model systems into consolidated architectures Peer‑reviewed publications or open‑source contributions in computer vision Experience in a fast‑growing SaaS or technology‑driven organization Version Française

À propos
du rôle Relevant du chef de la direction technologique (CTO), l'ingénieur principal ou l'ingénieure principale en apprentissage automatique est responsable de la recherche, du développement et du déploiement des modèles de vision par ordinateur au cœur de notre plateforme. Ceux-ci comprennent notamment les systèmes de segmentation qui détectent les défauts d'infrastructure et cartographient les actifs routiers à partir d'images acquises à grande échelle. Il s'agit d'un rôle technique pratique, axé sur la segmentation par transformeurs. Vous prendrez en charge l'ensemble du cycle de vie des modèles, de la recherche à la mise en production : amélioration de la précision et de la robustesse, généralisation des modèles à différents systèmes de caméras, consolidation de modèles spécialisés en architectures unifiées, et collaboration avec les équipes DevOps et de développement afin d'assurer leur déploiement fiable et à grande échelle. Responsabilités pri

📌 Senior Machine Learning Engineer (Winnipeg)
🏢 Citylogix
📍 Winnipeg