Research Scientist - Vision Language Model

What you'll do

• Develop VLMs across pre-training, instruction tuning, reasoning, and agents, including architectures and training methodologies.
• Build multimodal datasets, synthetic-data pipelines, and evaluation benchmarks; investigate capabilities such as OCR, grounding, document understanding, and visual question answering.
• Contribute to publications, technical reports, and open-source software; collaborate across teams and mentor junior researchers.

What they ask for

• Experience with large language models and/or vision-language models, including pre-training, fine-tuning, evaluation, or inference.
• Strong Python and PyTorch skills; experience with distributed training systems and large-scale model optimization.
• Familiarity with multimodal datasets and image, text, and video data pipelines; understanding of modern deep learning architectures, Transformers, attention, and multimodal fusion.
• Problem-solving, research, communication, and collaboration skills; ability to independently drive research or engineering projects.

Nice to have

• Hands-on experience training or fine-tuning large VLMs or multimodal foundation models at scale.
• Experience with distributed frameworks such as PyTorch Distributed, Megatron, Triton, or CUDA; multimodal reasoning, agentic systems, synthetic data, or efficient training and inference.

Pay

$150,000 – $450,000/year (base salary); USD 150,000–450,000 (per-year-salary; base salary)

Tools

Python, PyTorch, PyTorch Distributed, Megatron, Triton, CUDA, FlashAttention

AI focus

Research and development of multimodal foundation models, including visual understanding, language reasoning, tool use, agentic behavior, data-efficient training, and evaluation.