Senior Multimodal AI Engineer: Multimedia Pipelines

Meta; Madison (WI); On-site; USD 184,000 - 257,000; Full time

Job summary

Meta's Applied AI (AAI) organization is hiring a senior engineer to lead multimodal data pipelines for image, video, audio, and speech. You will define what data to produce, set measurement standards, and carry initiatives from exploration to production with cross-team impact.

You will mentor engineers, improve training efficiency, and build agentic workflows with human-in-the-loop design. This role requires deep experience in ML systems, Python, PyTorch, and large-scale data pipelines, with a

Qualifications

• Bachelor's degree in CS/CE or equivalent.
• 8+ years programming OR 4+ years with PhD.
• 3+ years building ML systems in production or research.
• Strong Python and PyTorch.
• Experience with speech, audio, or music ML (ASR, TTS, codecs, music IR or audio generation).
• Experience with large-scale data pipelines and distributed training.
• Track record of translating research ideas into working, measurable systems.

Responsibilities

• Set technical direction for a modality or capability area.
• Determine investment and define success metrics; explain tradeoffs.
• Drive direction across partner teams.
• Design and build agentic workflows with human-in-the-loop designs to automate data production.
• Design and own data pipelines: ingestion, filtering, pseudo-labeling, captioning, provenance tracking.
• Build evaluation infrastructure: objective metrics and human testing pipelines.
• Improve training efficiency: distributed training, GPU utilization.
• Reproduce and extend state-of-the-art research with ablations.
• Mentor engineers and contribute to hiring and onboarding.
• Build and train generative and representation models for speech, sound, and music.

Skills

Python

PyTorch

Multimodal ML

Distributed training

Data pipelines

Evaluation metrics

Education

Bachelor's degree in CS/CE or equivalent

Job description

Meta's Applied AI (AAI) organization is hiring a senior engineer to lead multimodal data pipelines for image, video, audio, and speech. You will define what data to produce, set measurement standards, and carry initiatives from exploration to production with cross-team impact.

You will mentor engineers, improve training efficiency, and build agentic workflows with human-in-the-loop design. This role requires deep experience in ML systems, Python, PyTorch, and large-scale data pipelines, with a