Reinforcement Learning Engineer

Bright Vision Technologies•Monroe Township, MI

1d•$100,000 - $150,000•Remote

About The Position

Bright Vision Technologies is seeking a Reinforcement Learning Engineer to design, build, evaluate, and deploy reinforcement learning solutions for sequential decision-making problems in real and simulated environments. The role includes developing simulation environments and scalable distributed training infrastructure, implementing modern RL methods, and improving training stability and sample efficiency. The engineer will develop reward functions and safety mechanisms, apply offline RL, imitation learning, and—when relevant—RLHF and related language-model fine-tuning techniques, and establish rigorous evaluation protocols. The position also involves monitoring deployed policies for drift, regression, and unintended behavior; collaborating with applied scientists and product teams; documenting technical decisions; and translating current research into production-ready solutions. This is a full-time, direct W-2 position that is 100% remote in the United States. The listed annual salary range is $100,000–$150,000. The employer is unable to sponsor new H-1B visa petitions, but encourages applications from U.S. citizens, Green Card holders, EAD holders, and H-1B transfer candidates.

Requirements

Master’s or PhD in Computer Science, Machine Learning, or a related field, or equivalent applied experience.
Ten or more years of combined RL research and engineering experience.
Strong proficiency in Python and modern deep learning frameworks.
Hands-on experience with at least one major RL library or in-house RL stack.
Solid understanding of probability, optimization, and the theoretical foundations of RL.
Experience designing and tuning reward functions in non-trivial environments.
Familiarity with simulation environments and large-scale experience collection.
Experience training neural network policies on GPU clusters.
Strong written and verbal communication skills.
Track record of shipping or publishing impactful RL work.

Nice To Haves

Experience with RLHF for large language models.
Familiarity with multi-agent RL or hierarchical RL.
Exposure to robotics, control systems, or autonomous driving.
Publications in RL or related research venues.
Open-source contributions to RL libraries or environments.

Responsibilities

Design and implement reinforcement learning solutions for sequential decision-making problems in real and simulated environments.
Develop, calibrate, and maintain simulation environments for large-scale agent training.
Implement and evaluate policy-gradient, actor-critic, off-policy, and offline RL algorithms.
Engineer reward functions and shaping strategies aligned with desired outcomes and safety constraints.
Apply offline RL and imitation learning when exploration is costly or unsafe.
Use RLHF, DPO, and related techniques to fine-tune large language models when relevant.
Build scalable distributed RL training infrastructure, including efficient experience collection and replay systems.
Improve training stability and sample efficiency through algorithmic and engineering changes.
Design rigorous evaluation protocols, including out-of-distribution and adversarial test cases.
Implement safety mechanisms, including constraint enforcement, conservative policies, and human-in-the-loop oversight.
Collaborate with applied scientists and product teams to identify high-value RL use cases.
Monitor deployed policies and models for drift, regression, and unintended behavior, and build alerting and dashboards to surface issues.
Document methodology, design decisions, and operational characteristics for internal stakeholders.
Stay current with RL research and translate promising techniques into production-ready solutions.