Machine Learning Engineer – Computer Vision

About Xantory

Role Overview

Computer Vision

• Specify camera selection and placement, lighting, capture schedules, image standards, and annotation standards for racks and trays.
• Build labelled image datasets from the company's crops, beginning with on-site data collection.
• Train and deploy models for plant detection, counting, growth stages, canopy coverage, stress, disease, and harvest readiness.
• Choose classification, detection, segmentation, or anomaly-detection approaches and validate performance on new crops and racks.
• Deploy inference to Raspberry Pi-class devices, Jetson-class devices where justified, and the site server.

Sensor and Outcome Data

• Combine camera results with temperature, humidity, CO₂, PPFD, pH, EC, flow, actuator history, stage boundaries, and yield data.
• Model cycle length and yield per rack against growing recipes and compare recipes using evidence.

Integration and Operation

• Ship models as services using Rust, PostgreSQL, Redpanda, Kubernetes, and Sentinel integrations.
• Version datasets, experiments, and models; monitor model performance and degradation; and retrain from production feedback.
• Define model data contracts with backend, edge, and console engineers.

Qualifications and Experience

• Three or more years of hands-on computer vision experience with PyTorch or TensorFlow is required.
• At least one vision system must have been taken from personal data collection through production operation.
• Experience building image datasets from scratch, including quality control, class balance, and handling small or shifting datasets, is required.
• Practical experience with cameras and video pipelines, exposure, colour under artificial light, and calibration is required.
• Horticulture, controlled-environment agriculture, plant phenotyping, or industrial quality inspection experience is advantageous.
• Multispectral or NIR imaging experience is advantageous.

Skills and Competencies

• Solid Python and machine learning fundamentals, including leakage-safe validation, are required.
• Production engineering experience with Git, code review, tests, containers, and services is required.
• Edge inference with ONNX, TensorRT, NVIDIA Jetson, or Raspberry Pi is advantageous.
• MLOps tooling experience is advantageous.
• Rust or Go, MQTT, and Kubernetes knowledge is advantageous.
• Clear written and spoken English is required.

Work Location

• The role is based in Dubai and requires in-person work with regular on-site farm activity.

Success Measures