Job Description

Responsibilities

- Apply relevant computer vision and machine learning techniques to solve challenges in depth estimation, semantic segmentation, and other vision tasks.

- Curate and manage diverse datasets, develop effective training strategies to boost model performance.

- Develop robust and well-optimized production models.

- Code using primarily Python and PyTorch. Conduct design and code reviews.

Minimum Qualifications

- 2+ years of industry experience in computer vision and deep learning, in areas such as segmentation, video understanding, and/or semi-supervised learning.

- Experience developing machine learning algorithms or infrastructure in Python and PyTorch.

- Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience.

- MS degree in Computer Science, Computer Vision, Machine Learning, or related technical field.

Preferred Qualifications

- Experience with 3D computer vision or computer graphics.

- Familiarity with ML model optimization.

- Experience developing production software with thoughtful code reviews and appropriate testing.

Additional Notes/Comments

Surrounding team & key projects

• XR platform core AI team. Have about 6 ICs, and the contractor will work closely with a subset of the 6 ICs.

• Main projects that this role will feed into is depth for mixed reality and virtual reality.

• The contractor will be working on one workstream within the project, and the workstream is large scale pre-training.

• Training ML models for depth for MR and VR applications.

Typical Day-to-Day in the role

• Work mostly in python and PyTorch.

• Manage using data from Instagram and Facebook. That could be writing scripts to select data, manage if the data is still public.

• Use the data to train depth models, figure out different ways to train the model. This will be the majority of their work.

• Link up the training to cases to show value or the impact of the model they have trained.

What makes this role interesting?

• Using state of the art, large scale, machine learning training strategies.

• Based off work that is recently published.

• At the forefront of computer vision machine learning.

How will performance be measured? Set up milestones, 3 stages of final deliverables/model output that the person can deliver, set up very distinct milestones for the candidate to follow along within the 3 stages.

Are there any types of candidate profiles or skills that may not be the right fit for this team? Candidates have trained models that are not vision related would not be the right fit.

Difficulties that the candidate should be aware they may face in the role and need to be able to handle to be successful in the role. Working with messy real-world data and having to produce a production model. Having to make sure the output model is robust for production.