Overview As an Applied Scientist in the NEST Science team, you will build and productionize computer vision models to turn visual data from Amazon's Middle Mile network into actionable decisions. You will own the scientific approach from problem framing to deployment, collaborating with product, operations, and engineering to address high-impact CV use cases. You'll work with large-scale, diverse image/video data and rigorous experiments to deliver measurable business impact. This role offers the chance toShape CV solutions for asset inspection, tracking, and condition monitoring at scale.
Compensation / Benefits- health insurance (medical, dental, vision)
- 401(k) matching
- paid time off
- parential leave
- RSUs
- sign-on payments
Responsibilities- Develop visual perception and representation learning across image/video domains (recognition, detection, segmentation, tracking) with advanced deep learning architectures
- Own large-scale model training, fine-tuning, and learning from heterogeneous or weakly supervised data (self-supervised, semi-supervised)
- Address real-world CV challenges at scale (unlabeled data, class imbalance, varied capture conditions)
- Collaborate with product/program, operations, and engineering stakeholders to translate problems into CV objectives with measurable criteria
- Design rigorous evaluation frameworks with precision/recall tradeoffs and business-cost aware operating points
- Build and maintain training/inference pipelines and drive production deployment with engineering teams
- Approach problems from first principles, selecting CV or alternative methods as appropriate
Key requirements- 3+ years of building ML models for business applications
- PhD, or Master's degree with 4+ years of CS/CE/ML-related experience
- Experience developing and implementing deep learning algorithms, especially in computer vision
- Proficiency in Java, C++, Python or related languages
- Experience in professional software development
- collaboration across cross-functional teams
- ability to frame ambiguous operational problems
- rigorous experimental design and data-driven decision making
- computer vision algorithms
- self-supervised / semi-supervised learning
- large-scale model training and deployment
