Overview As an Applied Scientist in Amazon Robotics, you advance mobile manipulation by developing learning-based methods for navigating and handling objects in dynamic fulfillment environments. You'll apply state-of-the-art research to production-scale robotics systems, collaborating with senior scientists and hardware teams to drive real-world impact. You will design architectures, train models, and build reusable data pipelines to improve robot performance across scenarios. This role combines research, engineering, and hands-on implementation to transform warehouse automation at scale. You will work in a fast-paced, collaborative environment with a strong focus on delivering measurable results.
Compensation / Benefits- Medical, Dental, and Vision Coverage
- Maternity and Parental Leave Options
- Paid Time Off
- 401(k) Plan
- Restricted Stock Units (RSUs)
Responsibilities- Design and train model architectures; iterate to optimize performance using diverse datasets
- Process and curate training data with governance, provenance, quality checks, and pipelines
- Design and run experiments in simulation and on physical embodiments; establish baselines and improve
- Write clean, well-documented code; contribute to training infrastructure and model evaluation tools
- Stay current with foundation models and robotics; contribute to literature, reports, and discussions
- Collaborate with senior scientists, engineers, and hardware teams; share knowledge and document best practices
Key requirements- 3+ years building models for business applications
- Master's degree (or PhD) in CS, CE, ML or related field with relevant experience
- Publications or patents in top-tier venues
- Proficient in Java, C++, Python or related languages
- Experience in algorithms, data structures, numerical optimization, data mining, parallel/distributed computing, HPC
- Experience using Unix/Linux
- cross-functional collaboration
- strong communication and documentation
- problem solving and initiative
- model development and training
- data governance and data pipelines
- experimentation and validation
