> Markdown version of [/jobs/ext/4690966-principal-platform-software-engineer](https://www.wearedevelopers.com/jobs/ext/4690966-principal-platform-software-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Principal Platform Software Engineer

- **Company:** Oracle
- **Location:** Seattle, WA, United States
- **Experience:** Expert
- **Salary:** $104,500 - $234,600
- **Contract:** Permanent contract
- **Skills:** 3D Scanning, Artificial Intelligence, Profiling, Software Debugging, Linux, Python (Programming Language), Kinematics, Machine Learning, Operational Data Store, Software Engineering, System Software, Reinforcement Learning, Graphics Processing Unit (GPU), Pytorch, Model Validation, Hardware Testing, Isaac Sim, Information Technology, Low Latency, TensorRT, Oracle Cloud Infrastructure, Software Library, Data Generation
- **Published:** November 12, 2026
- **Apply:** https://dejobs.org/x/x/B4FD35C256AB4FF1B367D3C1D9E83FEC/job/

## About the Role

* A bachelor's degree in Robotics, Computer Science, Mechanical Engineering, Electrical Engineering, or a related discipline, or equivalent practical experience.
 * At least six years of experience in machine learning engineering, software development, or applied robotics research, including substantial hands-on robot learning.
 * Hands-on experience training, fine-tuning, and evaluating robot policies using RL, SFT, and imitation learning, with measured task-level results.
 * Experience with robot foundation models, vision-language-action models, or other learned control policies. Examples include GR00T, ACT, and diffusion policies; comparable experience is equally relevant.
 * Practical understanding of RL environment and reward design, policy optimization, and learning diagnostics; ability to adapt pretrained models and train policies from demonstrations.
 * Strong Python and PyTorch skills, including training-loop implementation, GPU profiling, and debugging on Linux.
 * Hands-on development experience with NVIDIA Isaac Sim, including robot and scene configuration, physics, sensors, and automated experiments.
 * Experience curating compatible robot datasets, preventing leakage, designing reproducible experiments, and evaluating generalization and robot behavior.
 * Robotics fundamentals covering kinematics, dynamics, coordinate transforms, feedback control, and the effect of action representations, timing, and latency on learned behavior.
 * A record of selecting models and training methods under task, data, compute, and deployment constraints, owning delivery across components, and mentoring engineers., * Experience transferring learned policies to physical robots and diagnosing simulation-to-reality differences.
 * Experience with Isaac Lab, LeRobot, Gazebo, MuJoCo, or Nav2; synthetic data generation, 3D reconstruction, or domain randomization.
 * Experience optimizing robot-model inference with TensorRT or comparable tools under latency and memory constraints.
 * Experience with ROS 2, robot SDKs, URDF or USD assets, and model execution on robot or edge computers.
 * Experience scaling training across GPUs or contributing to robot-learning libraries.

## Description

Oracle Cloud Infrastructure (OCI) seeks a Principal Machine Learning Engineer for Robotics to lead simulation, policy training, fine-tuning, and evaluation using reinforcement learning (RL), supervised fine-tuning (SFT), and imitation learning.

The OCI Robotics Cloud team connects simulation, operational data, and AI models. You will turn demonstrations and operational feedback into compatible datasets, train policies, and evaluate robot behavior.

As an IC4 individual contributor, you will own technical direction across model selection, training, and evaluation, resolve ambiguous problems, and mentor engineers while remaining hands-on.

Responsibilities

What you will do

 * Select and adapt models and training approaches based on robotic tasks, available data, compute budgets, observation and action representations, and deployment constraints.
 * Develop SFT and imitation-learning workflows to adapt pretrained robot models and learn policies from demonstrations. Diagnose optimization failures, overfitting, and weak generalization.
 * Design RL environments and rewards in Isaac Sim, using Isaac Lab or comparable learning frameworks. Define observations, actions, resets, termination conditions, controllers, and curricula.
 * Train, tune, and debug RL policies, investigating unstable learning, unintended reward exploitation, and simulation or controller mismatches.
 * Prepare synchronized, compatible demonstration datasets. Validate sensor, state, action, timing, and normalization conventions; version training, validation, and evaluation splits and prevent leakage.
 * Lead baseline comparisons, ablations, and held-out evaluation. Measure task success, robustness, and generalization; establish model acceptance criteria and distinguish simulation results from hardware evidence.
 * Build reproducible GPU training and rollout workflows, tracking data, configurations, seeds, checkpoints, metrics, and recovery. Improve training throughput and resource use.
 * Use operational data to guide simulation calibration, domain randomization, and retraining. Partner with robotics engineers on inference performance, observation/action compatibility, and supervised hardware validation.
 * Lead architecture and technical reviews, resolve cross-team dependencies, mentor engineers, and deliver tested workflows. Validate AI-assisted code and experimental conclusions.