Applied Researcher: Continuous Learning for Deployed AI

Dyna Robotics; Redwood City (CA); On-site; USD 140,000 - 210,000; Full time

Job summary

Dyna Robotics is seeking an Applied Researcher to mine fleet sensor and video data, identify failure modes, and build monitoring that drives continuous model improvements in real-world deployments.

You will design end-to-end data pipelines, apply offline RL and reward modeling, and develop production monitoring to ensure day-one performance across sites. This is a hands-on, ship-it role with real impact.

Qualifications

• Hands-on experience with RL, sensor-data modeling, anomaly detection, or continual learning.
• Experience building monitoring, evaluation, or data pipelines for a live ML system.

Responsibilities

• Design and ship pipelines turning deployment data into targeted fine‑tuning and online policy improvements.
• Mine high-frequency multimodal sensor/video data to catch failure modes, drift, and regressions.
• Apply offline RL and reward modeling to improve policies from real-world deployment data.
• Build automated fleet monitoring that flags anomalies and near‑failures in real time.
• Characterize generalization gaps across new sites and develop evaluation harnesses.
• Turn findings into shipped improvements from data analyses to production dashboards and model updates.

Skills

Reinforcement learning

Sensor data modeling

Anomaly detection

Vision-language models

Continual/online learning

Python

PyTorch

JAX

Distributed compute

Education

Bachelor's/Master's/PhD in CS/Robotics/Statistics or related field

Tools

Slurm

GPU clusters

PyTorch

JAX

Job description

Dyna Robotics is seeking an Applied Researcher to mine fleet sensor and video data, identify failure modes, and build monitoring that drives continuous model improvements in real-world deployments.

You will design end-to-end data pipelines, apply offline RL and reward modeling, and develop production monitoring to ensure day-one performance across sites. This is a hands-on, ship-it role with real impact.