AI Trainer / Dataset Lead (YOLO Segmentation & ONNX) – Part-Time / Project Retainer

Location: Remote (Worldwide)

Company: Lariel Systems, LLC

Salary per gig job, salary posted is negotiable depending on skill.

About Us

Lariel Systems develops AI-powered engineering and infrastructure software. We are currently building road deterioration analysis systems using YOLO segmentation models and are looking for an experienced AI Trainer / Dataset Lead to help improve model performance and guide dataset quality.

This is not a full-time position. We are seeking a long-term project relationship or monthly retainer with someone who has practical experience building and improving computer vision datasets.

Responsibilities

- Review segmentation datasets and annotation quality.
- Analyze false positives and false negatives.
- Recommend new classes and hard-negative strategies.
- Develop labeling standards and annotation guidelines.
- Train and guide image annotators.
- Improve model accuracy through dataset refinement.
- Assist with YOLO training and validation.
- Export and validate ONNX models.
- Maintain dataset versions and quality documentation.

Required Skills

- Practical experience with YOLO (v8, v11, or similar).
- Experience with segmentation datasets.
- Understanding of hard negative mining.
- Familiarity with LabelMe, CVAT, or Label Studio.
- Experience with ONNX export and validation.
- Python familiarity.
- Strong communication and documentation skills.

Nice to Have

- Road infrastructure or pavement inspection experience.
- Experience with DirectML, ONNX Runtime, or OpenCV.
- Experience with Ultralytics YOLO.
- Knowledge of active learning workflows.

Engagement

- Part-time.
- Project-based or monthly retainer.
- Flexible schedule.
- Long-term opportunity as our platform grows.

To Apply

Please include:

1. A brief summary of your experience.
2. YOLO versions you have used.
3. Examples of segmentation projects you have worked on.
4. Experience with ONNX export.
5. Your preferred hourly or retainer rate.
6. Your time zone and weekly availability.

We value practical experience and problem-solving ability more than academic credentials. If you have successfully built and improved real-world datasets, we would love to hear from you.