The Model Quality Enablement (MQE) team is a group of Machine Learning (ML) experts within AI Infra - Trainers organization. Our unique strength is applying ML knowledge in the Infra domain to resolve challenges that affect model quality. Model Quality, commonly measured in Normalized Entropy (NE) at Meta, is one of the most important metrics for model training. It reflects the business impact a model can achieve. As such, it is critical to preventing model quality regression, which is common in enabling new training techniques or imposing resource constraints (e.g., training the same model with fewer GPUs). The MQE team works on a wide range of models in various projects, for instance:Build scalable solutions to root-cause, mitigate and prevent model quality regressions for various recommendation and ranking models..Enable end2end training of the new Sequential-based RecSys models and close the model quality gap between training inference models.Conduct applied research on developing novel training algorithms (e.g., Parameter Efficient Fine Tuning of LLMs) that maintain or improve model quality with various resource constraints.Our methodology is to develop fundamental model quality guardrails and optimization methods that are model type and training framework agnostic, rooting from a deep understanding of large scale model training/optimization process. The team collaborates closely with product groups including Ads (ML Infra and Core Ranking) and GenAI (MetaAI Search Engine). We also partner with other lower-level infra teams such as PyTorch and Performance & Capacity. The team offers great learning and growth opportunities for Infra/System generalists who want to develop ML skills, and vice versa for ML/AI generalists who want to learn about training infra and distributed systems.We are looking for PostDoc candidates who are interested in applied research on:You will have the opportunity to solve critical business problems at Meta by applying SotA techniques in the domain of LLMs, representation learning, RecSys and more. Publications to top Machine Learning conferences are encouraged and will be sponsored. This role will be a great entry point to experience doing applied research at an industry-leading company that created LLaMa.

• The application of LLMs in the domain of large-scale model training and productivity.
• The application of LLMs in Recommendation Systems.
• Novel training techniques for LLMs and Recommendation Systems with resource constraints.

• Currently has or is in the process of obtaining a PhD degree or completing a postdoctoral assignment in Computer Science or a similar field. Degree must be completed prior to joining Meta.
• Research experience in NLP, Large Language Models, RecSys.
• Experience with PyTorch and distributed machine learning.
• Programming experience using Python and PyTorch.
• Currently has, or is in the process of obtaining a Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience. Degree must be completed prior to joining Meta.
• Must obtain work authorization in country of employment at the time of hire and maintain ongoing work authorization during employment.

• Demonstrated research and software engineering experience via an internship, work experience, coding competitions, or widely used contributions in open source repositories (e.g. GitHub).
• Experience solving complex problems and comparing alternative solutions, tradeoffs, and diverse points of view to determine a path forward.
• Experience working and communicating cross functionally in a team environment.