AI Research Engineer (Agentic Reinforcement Learning) - 100% Remote Worldwide

Tether Operations Limited

9h • Remote

About The Position

As a member of Tether’s AI model team, the AI Research Engineer will advance multimodal reinforcement learning for next-generation intelligent systems. The work focuses on improving decision-making and adaptive behavior across modalities such as text, images, and audio, including for real-world and domain-specific challenges. The role involves research and development across diffusion-based generative models, autoregressive models for multimodal understanding and reasoning, unified multimodal frameworks, and resource-efficient models for constrained hardware.

The engineer will build reinforcement learning infrastructure and reward-modeling strategies for efficient, stable, large-scale training, while addressing reward hacking and other failure modes. The role also involves curating simulation environments and datasets, improving policy performance across modalities, finding bottlenecks in multimodal learning and reward optimization, and exploring approaches that learn more effectively from environmental feedback. Research findings are expected to be published in top-tier conferences.

Requirements

• A Master's degree in Computer Science or a related field.
• Proven experience running large-scale reinforcement learning experiments in multimodal and vision-centric systems, including online RL settings, with demonstrated impact on domain-specific decision-making and measurable improvements in policy performance.
• Deep understanding of reinforcement learning algorithms and optimization methods for vision and multimodal learning, including policy stability, exploration, and sample efficiency in complex, high-dimensional environments involving images, video, and other modalities.
• Strong proficiency in PyTorch and deep learning frameworks for vision and multimodal AI.
• Hands-on experience building end-to-end reinforcement learning pipelines covering simulation, training, evaluation, and deployment in production-grade systems.
• Demonstrated ability to apply empirical research to multimodal and vision RL challenges, including sample inefficiency, exploration-exploitation tradeoffs, and training instability.
• Experience designing robust evaluation frameworks and iterating on algorithmic improvements to advance agent performance.
• A strong track record of AI research and publications in top-tier conferences.

Nice To Haves

• A PhD in Machine Learning, NLP, Computer Vision, or a closely related discipline.

Responsibilities

• Conduct research on reinforcement learning algorithms for multimodal models, including diffusion-based approaches, autoregressive models for multimodal understanding, and unified frameworks integrating multiple modalities.
• Design and build efficient, reliable infrastructure for scalable, distributed training across multimodal systems.
• Develop and refine reward-modeling strategies to improve training stability and align model behavior with desired outcomes while mitigating reward hacking and related failure modes.
• Create and curate multimodal simulation environments and datasets for training, evaluation, and benchmarking.
• Design and conduct benchmarking and evaluation protocols to measure performance, track progress against baselines, and validate improvements across multimodal tasks.
• Analyze and optimize policy performance across modalities by identifying bottlenecks in training, credit assignment, and cross-modal alignment.
• Investigate and develop reinforcement learning paradigms that learn more effectively from environment feedback and improve performance in dynamic, real-world environments.
• Publish research findings in top-tier conferences such as ICML, NeurIPS, ICLR, CVPR, ICCV, and ECCV.

What This Job Offers

• Job Type: Full-time
• Career Level: Mid Level