RL Engineer, End-to-End Autonomy
ExternalPrepare for this interview
EliteAI-generated questions, company research, and talking points tailored to this role
About the role
Our mission is to bring the power of E2E machine learning and robotics to John Deere, revolutionizing how robotic systems are built. We aim to open up previously inaccessible opportunities, delivering autonomy products that customers love. As an RL Engineer, you will help grow and shape our E2E stack. This involves shaping the design of the E2E training and inference pipeline, both on- and off-vehicle and on- and off-prem. You will also help with new models and features. Our E2E team is lean and moving quickly. You should be comfortable and excited about working with ambiguity, helping define what will move the program forward, working across traditional boundaries, and learning new things. Employment Type : Full-Time Work Location : Santa Clara, CA (expect 3 days in the office) Visa sponsorship is available for this position. Job Responsibilities The main job responsibilities include: Design and implement state-of-the-art online and offline reinforcement learning algorithms for complex tasks with long horizons. Formulate reward models and exploration strategies that enable task performance and adherence to strict safety requirements. Enable flexible customer operation across a range of tasks with natural language instructions. Enhance policy robustness to challenges such as sensor noise, machine wear, and extreme environmental variability. Design and conduct experiments; develop evaluation frameworks for simulation and real-world deployment. Collaborate closely with infrastructure engineers to design scalable E2E training systems, including large-scale simulation infrastructure. Collaborate with pretraining and robotics engineers to integrate RL seamlessly into the full E2E autonomy stack. Drive the E2E vision and act as an ambassador for an E2E-first organization. Technical leadership and mentorship: serve as an in-house RL expert, elevating the team through code reviews, algorithmic guidance, and fostering a culture of rigorous scientific experimentation. Stay up to date with the latest RL research and integrate advancements into our stack. Required Experience and Skills Proven track record in developing RL models. Prefer experience deploying to production for robotics and boosting key metrics. Expertise in developing simulation environments and tackling the sim-to-real transfer. Expertise in designing and developing software for complex systems. Comfortable working on new hardware systems and working on new RL/ML/software problems. Strong Python coding skills and proficiency with deep learning frameworks like PyTorch. Comfortable working across traditional team boundaries to deliver results. Excellent brainstorming, creative thinking, mathematical analysis, and communication skills. Track record of regularly anticipating technical issues and making architectural and design decisions to avoid them. BS/MS/PhD in CS or related field, and 3+ years delivering high-performance RL products professionally. Preferred Experience and Skills Experience with robotics middleware such as ROS or other robotics-focused software packages. Strong CUDA background or other GPU frameworks. Experience in multidisciplinary environments. (We've got CS, CV/ML, EE, ME, etc.) Have worked on embedded systems. Experience with system architecture. At Blue River, we're passionate about creating an inclusive workplace that promotes and values diversity. While we have more work to do to advance diversity and inclusion, we're investing in our programs, including recruiting, mentorship, career development, and learning & development to ensure they support our Diversity, Equity, and Inclusion goals. We support each employee in living a full life, enabling a thriving career, and accomplishing a meaningful, challenging mission while collaborating with incredible people. We are dedicated to building a diverse and inclusive workplace, so if you're excited about this role but your experience doesn't align completely