AI Research Engineer (Pre-training - LLM & Multi-Modal)

Tether is a global fintech leader building reserve-backed digital assets and blockchain infrastructure that power hundreds of millions of users worldwide. The company operates across multiple verticals including stablecoins, energy solutions, peer-to-peer communications technology, and digital education. You'll join their Data team as an AI Research Engineer focused on large-scale pre-training for language models and multi-modal systems.

Tether has established itself as a trusted name in the Web3 and fintech space by combining technological innovation with transparent operations. Their commitment to building lean, high-performing teams across remote locations has made them a destination for top technical talent globally. The company's expansion into AI reflects the broader 94% surge in AI/ML engineering vacancies and the growing demand for researchers who can scale models across distributed infrastructure.

What you'll do

• Lead foundational pre-training initiatives for large language models and multi-modal architectures (combining text, vision, audio, and other modalities) on distributed servers with thousands of NVIDIA GPUs across multiple nodes
• Design and prototype novel architectures, tokenizers, and cross-modal alignment layers that strengthen model intelligence and multi-modal reasoning capabilities
• Build comprehensive data pipelines to source, filter, and curate massive-scale textual and multi-modal datasets for efficient pre-training at scale
• Execute experiments independently and collaboratively, analyzing results to refine training methodologies and optimize token efficiency
• Debug and resolve performance bottlenecks in model efficiency, computational performance, and multi-modal alignment stability during extended training runs
• Advance distributed training systems to improve scalability and hardware utilization across target platforms

What you'll bring

• A degree in Computer Science or related field, ideally with a PhD in NLP, Machine Learning, or adjacent discipline and a strong publication record at top-tier conferences
• Hands-on experience running large-scale LLM or multi-modal pre-training on distributed GPU clusters with thousands of NVIDIA processors
• Practical expertise with distributed training frameworks, libraries, and tools at production scale
• Deep knowledge of transformer and non-transformer architectures designed to enhance intelligence, efficiency, and scalability
• Advanced proficiency in PyTorch and Hugging Face libraries with demonstrated experience in model development, continual pre-training, and deployment
• Strong English communication skills and a research-driven mindset with ability to explore novel techniques and algorithms

This is a fully remote position offering you the opportunity to work alongside a globally distributed team of elite technical researchers pushing the boundaries of AI capabilities.