Research Scientist Intern, Muse Post-Training & Flywheel (PhD)
Generate a McCoy IQ challenge in 30 seconds.
See how candidates think and approach the work this role demands, before the phone screen. We'll build a video challenge from this posting, and you can edit or share it before it goes live.
Key details
Why this role is notable
Meta is institutionalizing a closed loop that turns live product feedback and hallucination failures directly into post-training data and evaluations for its Muse models, treating user interaction as a continuous training signal rather than a one-off fine-tuning step.
Job Description
Meta is seeking PhD interns with strong machine learning and programming skills to improve Muse through applied product projects. Interns will work on the post-training flywheel, turning product feedback into measurable model improvements.
Responsibilities
Improve pipelines connecting product feedback, evaluation, training data, and Muse post-training. Build evaluations and curate training data to address product failures, including hallucinations. Run training experiments and work with researchers and engineers to improve model performance on real user tasks.
Qualifications
Currently pursuing a Ph.D. in Computer Science, Machine Learning, Natural Language Processing, or a related field Programming experience in Python and deep learning frameworks such as PyTorch Hands-on experience with model training, evaluation, or training data pipelines Must obtain work authorization in the country of employment at the time of hire and maintain ongoing work authorization during employment Experience with large language model post-training, reinforcement learning, or training data curation Intent to return to the degree program after completing the internship
