AI Research Engineer - Reinforcement Learning (100% remote Worldwide)

Tether Operations Limited

4d • Remote

About The Position

As a member of the AI model team, you will drive innovation in reinforcement learning approaches for advanced models. Your work will optimize decision-making and adaptive behavior to deliver enhanced intelligence, improved performance, and domain-specific capabilities for real-world challenges. You will work across a broad spectrum of systems, including resource-efficient models designed for limited hardware environments and complex multi-modal architectures that integrate data such as text, images, and audio.
We expect you to have deep expertise in designing reinforcement learning systems and a strong background in advanced model architectures. You will adopt a hands-on, research-driven approach to developing, testing, and implementing novel reinforcement learning algorithms and training frameworks. Your responsibilities include curating specialized simulation environments and training datasets, strengthening baseline policy performance, and identifying as well as resolving bottlenecks in the reinforcement learning process. The ultimate goal is to unlock superior, domain-adapted AI performance and push the limits of what these models can achieve in dynamic, real-world environments.

Requirements

• A degree in Computer Science or related field. Ideally PhD in NLP, Machine Learning, or a related field, complemented by a solid track record in AI R&D (with good publications in A conferences).
• Proven experience with large-scale reinforcement learning experiments, including online RL techniques such as Group Relative Policy Optimization (GRPO), is essential. Your contributions should have led to measurable improvements in domain-specific decision-making and overall policy performance.
• Deep understanding of reinforcement learning algorithms is required, including state-of-the-art online RL methods and other gradient-based optimization approaches like policy gradients, actor-critic, and GRPO. Your expertise should emphasize enhancing policy stability, exploration, and sample efficiency in complex, dynamic environments.
• Strong expertise in PyTorch and relevant reinforcement learning frameworks is a must. Practical experience in developing RL pipelines, from simulation and online training to post-training evaluation and deploying RL-based solutions in production environments is expected.
• Demonstrated ability to apply empirical research to overcome reinforcement learning challenges such as sample inefficiency, exploration-exploitation tradeoffs, and training instability. You should be proficient in designing robust evaluation frameworks and iterating on algorithmic innovations to continuously push the boundaries of RL agent performance.

Responsibilities

• Develop and implement state-of-the-art reinforcement learning algorithms designed to optimize decision-making processes in both simulated and real-world settings. Establish clear performance targets such as reward maximization and policy stability.
• Build, run, and monitor controlled reinforcement learning experiments. Track key performance indicators while documenting iterative results and comparing outcomes against established benchmarks.
• Identify and curate high-quality simulation environments and training datasets that are tailored to specific domain challenges. Set measurable criteria to ensure that the selection and preparation of these resources significantly enhance the learning process and overall model performance.
• Systematically debug and optimize the reinforcement learning pipeline by analyzing both computational efficiency and learning performance metrics. Address issues such as reward signal noise, exploration strategy, and policy divergence to improve convergence and stability.
• Collaborate with cross-functional teams to integrate reinforcement learning agents into production systems. Define clear success metrics such as real-world performance improvements and robustness under varied conditions and ensure continuous monitoring and iterative refinements for sustained domain adaptation.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level
• Education Level: Ph.D. or professional degree