Foundation AI Engineer — LLMs, Fine-Tuning & Scale

XVILABS, INC.; Cambridge (MA); On-site; USD 140,000 - 210,000; Full time

Job summary

XVI Labs in Cambridge, MA is seeking a research engineer to develop end-to-end workflows for foundation models, from data preparation through fine-tuning and evaluation. You will design experiments across architectures and objectives, tracking results to improve model quality.

Collaborate with engineers to scale GPU training, deploy models, and optimize inference. Proficiency in Python, PyTorch, and the Hugging Face tools is essential; experience with RL and fine-tuning LLMs is highly valued.

Qualifications

• Strong Python skills and hands-on experience with PyTorch and the Hugging Face ecosystem, including Transformers and Datasets.
• Experience with deep learning and reinforcement learning, with a solid understanding of neural networks, transformers, training objectives, and model evaluation.
• Experience training or fine-tuning LLMs or other foundation models, with the ability to debug training runs and explain experimental results.

Responsibilities

• Develop pre-training and post-training workflows for foundation models, including LLMs, from data preparation through fine-tuning and evaluation.
• Design experiments across model architectures, training objectives, and datasets; track results and improve model quality.
• Work with engineers to scale GPU training, deploy models, and improve inference performance and reliability.

Skills

Python

PyTorch

Hugging Face Transformers

Hugging Face Datasets

Deep learning

Reinforcement learning

LLMs / foundation models

Tools

Distributed training

Job description

XVI Labs in Cambridge, MA is seeking a research engineer to develop end-to-end workflows for foundation models, from data preparation through fine-tuning and evaluation. You will design experiments across architectures and objectives, tracking results to improve model quality.

Collaborate with engineers to scale GPU training, deploy models, and optimize inference. Proficiency in Python, PyTorch, and the Hugging Face tools is essential; experience with RL and fine-tuning LLMs is highly valued.