AI Research Engineer

AI Research +1Updated on Nov 11

Full-timeNo salary listedIn-person

About the Job

Requirements

The role is aimed at experienced engineers, but no specific years of experience, degree, or required qualification is stated.

Responsibilities

Build post-training, evaluation, and serving systems that turn frontier large language models into reliable, high-quality product experiences.
Build a general-purpose agentic platform that powers generative AI products and enables teams to ship faster, more safely, and at scale.
Build systems to optimize model capacity and cost through fine-tuning, post-training, and other techniques.
Adapt and scale systems across Meta products.
Own work from training data through production behavior and measurement.
Develop post-training pipelines, including supervised fine-tuning, reinforcement learning from human feedback, and synthetic data generation.
Develop evaluation methodologies, including auto-judge design, benchmark construction, and human-AI calibration.
Build production serving systems involving retrieval-augmented generation, memory, and multimodal generation.
Work on multi-agent orchestration and end-to-end agentic product experiences, shipping to real users at scale.

We’re hiring Research Engineers to join teams across Meta working at the intersection of frontier AI and real-world product impact. You’ll be embedded directly in Facebook’s ecosystem, helping reimagine core experiences and reshape how people discover content, connect with creators, and interact with each other. The work spans some of the most bold bets in applied GenAI, including: - Building the post-training, evaluation, and serving systems that turn frontier LLMs into reliable, high-quality product experiences used by billions. - Building a general-purpose agentic platform that powers a wide range of GenAI products across Facebook —enabling teams to ship faster, safer, and at scale. - Building systems that enable capacity and cost optimizations through model fine-tuning, post-training and other techniques. - Adapting and scaling these systems across Meta’s products. Why Join Us - Product LLM work at singular scale Your post-training decisions, evaluation frameworks, and serving architecture directly affect billions of daily interactions. - End-to-end ownership We don't hand off models to a separate product team. We own the loop from training data to production behavior to measurement. The impact of your work shows up in days, not quarters. - The problems are unsolved How do you evaluate open-ended conversational AI at scale? How do you fine-tune for groundedness across millions of varied creator profiles? These aren't incremental improvements, they're open research questions with immediate product consequences. - Our team is hands-on, with high autonomy, working on critical bets We're deliberately keeping this team lean and experienced. You'll have outsized influence on technical direction, not just execution. Depending on your interests and strengths, your work could span post-training pipelines (SFT, RLHF, synthetic data generation), evaluation methodology (auto-judge design, benchmark construction, human-AI calibration), production serving systems (RAG, memory, multi-modal generation), multi-agent orchestration, or E2E experience of building agentic products, - all grounded in shipping to real users at scale

About the Company

Recently Posted Jobs

Explore more opportunities at Meta

See more jobs