Tether.to is a digital asset company focused on stablecoin infrastructure and blockchain-based payments. Its Tether tokens are designed to maintain a 1-to-1 value with corresponding fiat currencies and can be transferred across multiple blockchains, supporting digital transactions and cross-border payments. The company also publishes information about its assets and reserves and develops educational initiatives intended to expand understanding and use of digital assets, including in the Middle East, Turkey, and the Philippines. With its work spanning crypto, fintech, and finance, Tether.to contributes to the development of alternative financial infrastructure and stable digital currency applications.
AI Research Engineer, Model Compression and Quantization
Develop compression methods for large language and vision-language models used in Tether’s blockchain-based digital finance platform. Improve quantization, distillation, and pruning to run generative AI efficiently on edge devices.
Description
- Reduce generative AI model size and inference latency with low-bit quantization while preserving accuracy and output quality.
- Distill larger teacher models into smaller student models for multimodal reasoning across text, images, and audio.
- Prune redundant model parameters and attention heads.
- Evaluate trade-offs among model size, latency, memory, throughput, and accuracy.
- Research and apply mixed-precision quantization and other advanced compression methods.
- Track research on compressing multimodal and generative model architectures.
- Build reliable compression pipelines and define performance and fidelity metrics.
- Resolve bottlenecks in production inference.
- Develop scalable AI systems for edge devices such as smartphones, with low memory use and latency.
- Record methods, experiments, and results so work can be reproduced and shared.
- Write technical papers and publish findings at conferences such as NeurIPS, ICML, ICLR, CVPR, ACL, and AAAI.
Requirements
- Hold a degree in computer science or a related discipline.
- A PhD in NLP, machine learning, or a related discipline is preferred.
- Demonstrate a strong AI research and development record, including publications at leading conferences.
- Have experience with PyTorch or a comparable deep learning framework.
- Bring practical experience in quantization-aware training and post-training quantization.
- Have research and practical experience with knowledge distillation.
- Have research and practical experience with model pruning.
- Understand neural network architectures and training, including transformers, LLMs, VLMs, backpropagation, optimization, and fine-tuning.
- C++ familiarity is an advantage.
- Communicate effectively in English.
- Specify the country where you intend to work.
Benefits
- Work remotely from anywhere in the world.
- Collaborate with a global team and experts across fintech, AI, and blockchain.
- Contribute to digital finance, blockchain, AI, and peer-to-peer technology products.
- Have the opportunity to write technical papers and publish at leading conferences.
- Access professional and research collaboration opportunities.