Software Engineer, Multimedia & Multimodal AI
Core
Design and build agentic, human-in-the-loop, and expert-in-the-loop workflows and pipelines for scalable multimedia data production, including ingestion, filtering, pseudo-labeling, captioning, and audio-corpus provenance tracking.
Role type
Senior IC machine-learning engineer (multimedia & multimodal AI)
Builds
Large-scale data pipelines, evaluation infrastructure, and generative/representation models for speech, sound, and music
Domain
Multimedia AI, audio generative modeling, speech processing
Deliverable
production ML models
Required skills
Python, PyTorch, large-scale data pipelines, distributed training, speech/audio/music machine learning, generative modeling, representation learning, experiment management
Preferred skills
publications at major venues, continuous-data generative modeling, audio DSP, controllable generation, human-evaluation pipelines, music expertise, multimedia data lineage, AI agents, multimodal-model evaluation, responsible AI, zero-to-one process building, workflow optimization with AI tools
Technologies
PyTorch, Python
Responsibilities
Design and build agentic, human-in-the-loop, and expert-in-the-loop workflows and pipelines for scalable multimedia data production; Own large-scale data pipelines covering ingestion, filtering, pseudo-labeling, captioning, and audio-corpus provenance tracking; Build evaluation infrastructure including objective metrics, human listening tests, and correlation analysis between automated and human evaluations; Improve distributed training efficiency and reliability, GPU utilization, codec and tokenizer retraining, and experiment management; Implement and extend state-of-the-art research methods, conduct rigorous ablations, and translate research ideas into measurable systems; Build and train generative and representation models for speech, sound, and music, including text-, audio-, and video-conditioned generation, infilling, editing, and style transfer; Mentor engineers, contribute to hiring and onboarding, and establish quality and evaluation standards
