Research Engineer - AI Trust - Meta Superintelligence Labs

Meta Meta · Big Tech · Menlo Park, CA +1

Research Engineer focused on AI safety for large language and multimodal models, involving designing, implementing, and evaluating safety techniques, curating datasets, fine-tuning models, and building scalable infrastructure for safety evaluation and mitigation. Requires experience with LLM training, fine-tuning, evaluation, and safety research, along with Python and PyTorch.

What you'd actually do

  1. Design, implement, and evaluate novel, systemic, and foundational safety techniques for large language models and multimodal AI systems
  2. Create, curate, and analyze high-quality datasets for safety system and foundations
  3. Fine-tune and evaluate LLMs to adhere to Meta’s safety policies and evolving global standards
  4. Build scalable infrastructure and tools for safety evaluation, monitoring, and rapid mitigation of emerging risks
  5. Work closely with researchers, engineers, and cross-functional partners to integrate safety solutions into Meta’s products and services

Skills

Required

  • Python
  • PyTorch
  • LLM/NLP
  • computer vision
  • AI/ML model training
  • technical lead experience
  • complex technical projects

Nice to have

  • PhD in Computer Science, Machine Learning, or a relevant technical field
  • multilingual LLM experience
  • multimodal AI experience (text, image, voice, video, reasoning)
  • distributed training of LLMs
  • scalable safety mitigations
  • automation of safety tooling

What the JD emphasized

  • Publications at peer-reviewed conferences (e.g. ICLR, NeurIPS, ICML, KDD, CVPR, ICCV, ACL)
  • Experience developing, fine-tuning, or evaluating LLMs across multiple languages and capabilities (text, image, voice, video, reasoning, etc)
  • Demonstrated experience to innovate in safety foundational research, including custom guideline enforcement, dynamic policy adaptation, and rapid hotfixing of model vulnerabilities
  • Experience designing, curating, and evaluating safety datasets, including adversarial and borderline prompt cases
  • Experience with distributed training of LLMs (hundreds/thousands of GPUs), scalable safety mitigations, and automation of safety tooling

Other signals

  • foundational safety techniques
  • large language models
  • multimodal AI systems
  • safety evaluation
  • policy adherence