Applied Scientist Intern - Trust and Safety (Multimodal Foundation Model) - Global Frontier Tech Recruitment Program - 2027 Start (PhD)

What you'll do

• Research and build multimodal moderation models, including MoE training, cross-modal reasoning, and unified understanding and generation.
• Develop agentic moderation systems using reinforcement learning, multi-step decision-making, and tool collaboration.
• Improve moderation systems’ generalization and adversarial robustness across languages and content types.

What they ask for

• Currently pursuing a PhD in Computer Science, Data Science, Artificial Intelligence, or a related field.
• Proficiency in Python, Rust, or C++ and experience with deep learning frameworks.
• Strong understanding of distributed computing frameworks and performance tuning and verification for training, fine-tuning, or inference.

Nice to have

• Published research papers or contributions to the LLM community.
• Experience with inference acceleration, AI system evaluation, LLM applications or agents, or technologies such as PEFT, RL, MoE, CoT, or Langchain.

Tools

Python, Rust, C++, pytorch, deepspeed, megatron, vllm, GRPO/PPO

AI focus

Research and develop multimodal moderation foundation models and RL-driven agentic moderation, including MoE architectures, cross-modal reasoning, synthetic data generation, tool use, and adversarial robustness.