Computer Vision Engineer

About the role
As a Computer Vision Engineer at Orom, you will work across the stack to develop, deploy, and improve computer vision algorithms that operate reliably in the real world. This role sits between research, engineering, and production: you will help design novel perception algorithms, understand why systems fail in deployment, and turn those insights into robust, maintainable solutions.

You will work on problems involving 3D geometry, visual perception, machine learning, and real-world sensing. Your work will not stop at offline experiments or benchmark performance. You will be expected to reason about production failures, debug edge cases, improve model and algorithm behavior under messy real-world conditions, and collaborate closely with ML, robotics, C++, infrastructure, and product-facing teams.

This role is for someone who is comfortable moving between Python and/or C++, research code and production systems, algorithm design and practical debugging. You should care about building computer vision systems that are technically strong, explainable, reliable, and useful in deployed environments.

Responsibilities

• Design, develop, and improve computer vision algorithms for real-world perception, 3D understanding, spatial reasoning, and scene interpretation.
• Work across the stack, using Python and/or C++ to prototype, evaluate, optimize, and productionize perception algorithms.
• Develop novel approaches that combine machine learning, 3D geometry, classical computer vision, and practical engineering constraints.
• Investigate production failures by analyzing model outputs, sensor data, edge cases, domain shifts, and real-world operating conditions.
• Improve algorithm robustness across lighting changes, motion, occlusions, viewpoint variation, sensor noise, calibration issues, and messy deployment data.
• Build evaluation tools, metrics, and debugging workflows that expose failure modes and reflect real deployment performance rather than only benchmark accuracy.
• Collaborate closely with ML researchers, C++ engineers, robotics engineers, and MLOps to integrate computer vision algorithms into production systems.
• Contribute to data collection, data curation, failure mining, and retraining workflows that help perception systems improve over time.
• Translate prototype algorithms into reliable, maintainable, and efficient implementations that can run within product and hardware constraints.
• Support monitoring and analysis of deployed perception systems, including identifying regressions, diagnosing root causes, and proposing practical fixes.
• Communicate technical tradeoffs clearly across research, engineering, and product teams, especially when balancing accuracy, latency, reliability, explainability, and maintainability.
• Contribute to technical documentation, implementation plans, experiment reports, and clear analysis of algorithm limitations.

What we are looking for

• Proficiency in Python and/or C++, with the ability to write clean, reliable code for experimentation, tooling, and production integration.
• Strong experience in computer vision, visual perception, 3D geometry, machine learning, robotics, or a related technical field.
• Solid understanding of the fundamentals of 3D geometry, including camera models, transformations, projections, calibration, pose estimation, triangulation, or reconstruction.
• Strong practical knowledge of machine learning and modern computer vision workflows, including model development, evaluation, debugging, and error analysis.
• Ability to move between algorithm design, implementation, experimentation, debugging, and production constraints.
• Experience building or improving perception systems that operate on real-world data, not only curated datasets or academic benchmarks.
• Strong debugging skills and the ability to reason from observed failures back to data, model, algorithm, or system-level root causes.
• Familiarity with classical computer vision and/or learning-based perception methods, and the judgment to choose the right tool for the problem.
• Comfortable collaborating closely across teams and disciplines, especially with ML, robotics, C++, infrastructure, and product-facing engineers.
• Clear communication skills and a pragmatic approach to technical uncertainty, tradeoffs, and iterative improvement.

Bonus points for experience with robotics, autonomous systems, spatial computing, mapping, localization, or real-time perception systems, particularly involving SLAM, structure-from-motion, 3D reconstruction, tracking, calibration, or multi-view geometry.