Remote AI Research Engineer - Model Compression

Tether Ltd.; Dublin; Hybrid; EUR 90.000 - 130.000; Vollzeit

Zusammenfassung

Tether Ltd. is building cutting-edge AI tech for multimodal models and efficient deployment on edge devices. You will drive compression research for LLMs and VLMs, focusing on quantization, distillation, and pruning to achieve high fidelity with low resource requirements.

You will collaborate across teams, publish findings, and push frontier research in model compression while balancing accuracy, latency, and memory constraints. A PhD and strong PyTorch background are highly valued for this role.

Qualifikationen

• PhD in Computer Science or related field; NLP/ML PhD preferred.
• Proven experience with PyTorch.
• Hands-on experience with model quantization (QAT and PTQ).
• Research and hands-on experience with knowledge distillation for compressing large models.
• Experience with model pruning for compressing large models.
• Strong understanding of transformers architectures (LLMs/VLMs).
• Familiarity with C++ is a plus for low-level optimization.

Aufgaben

• Develop and implement compression pipelines to reduce model size and latency.
• Advance quantization, distillation, and pruning techniques for multimodal AI systems.
• Quantify trade-offs between size, latency, memory, and accuracy; propose improvements.
• Publish findings and document methodologies for reproducibility and collaboration.
• Stay current with NeurIPS, ICML, ICLR, CVPR, ACL, AAAI and related venues.

Kenntnisse

PyTorch

Transformers

Knowledge distillation

Model pruning

Quantization

C++ familiarity

Ausbildung

PhD in NLP / ML

CS related degree

Jobbeschreibung

Tether Ltd. is building cutting-edge AI tech for multimodal models and efficient deployment on edge devices. You will drive compression research for LLMs and VLMs, focusing on quantization, distillation, and pruning to achieve high fidelity with low resource requirements.

You will collaborate across teams, publish findings, and push frontier research in model compression while balancing accuracy, latency, and memory constraints. A PhD and strong PyTorch background are highly valued for this role.