Responsibilities
- Design and implement models that capture temporal dynamics for recognizing events, states, and detailed action sequences over time.
- Evaluate proprietary models relative to current benchmarks and propose optimal perception system architectures.
- Specify inputs and outputs for perception systems and validate their feasibility using historical recorded data.
- Build a continuous feedback system incorporating discrepancy detection, active learning, and human-assisted relabeling, along with tools and interfaces for offline testing and real-time readiness.
- Collaborate with data labeling and data management teams to refine annotation schemas, ensure quality control, and optimize data pipelines for machine learning.
- Bridge the gap between offline model validation and real-time system integration by establishing a sustained human-in-the-loop data refinement process.
- Work closely with AI researchers, roboticists, data engineers, and cross-functional partners to deliver a flexible perception system supporting fast experimentation and product development.
Compensation
Competitive salary and benefits package
Work Arrangement
Hybrid work model with flexibility for remote and on-site collaboration
Team
Collaborative environment integrating AI, robotics, and data science to advance intelligent systems
Responsibilities
- Develop temporal models for activity and workflow understanding: event/state recognition and fine-grained temporal action segmentation.
- Benchmark in-house models against the state of the art and recommend the target perception architecture.
- Define the perception input/output specification and demonstrate offline feasibility on recorded data.
- Stand up a continuous-improvement loop (discrepancy flagging, active learning, human-in-the-loop relabeling) and the tooling/UI needed for offline evaluation and the path to real-time use.
- Partner with annotation and data teams to shape label taxonomies, QC, and the data pipeline that feeds the AI/ML models.
- Establish the path from offline evaluation on recorded data to real-time integration, including the continuous-improvement (human-in-the-loop) data loop.
- Partner with AI/ML researchers, robotics, data engineers, and other stakeholders to deliver a perception layer that enables rapid prototyping and learning while working toward a product solution.
Available for qualified candidates
