Senior Multimodal ML Engineer — Vision & Audio for Production

White Circle; Paris; Hybride; EUR 122 613 - 201 436; Plein temps

Avantages offerts par ce poste

Relocation package

Hybrid work from Paris

Medical insurance (France)

All the hardware, tools & services

Subscriptions for AI agents & IDEs

Team off-sites

Résumé du poste

White Circle is hiring a Multimodal ML Engineer in Paris. You will train and ship vision, audio, video, and speech models for an AI safety platform processing 100M+ API calls monthly.

You will extend models across modalities, design experiments, and build multimodal data pipelines for training-ready datasets, including synthetic data generation. You will deploy models end-to-end, define evaluation benchmarks, and optimize for production latency.

Qualifications

• 3+ years training large-scale deep learning models in multimodal domains.
• Strong PyTorch skills with hands-on distributed training experience (DeepSpeed, FSDP, or similar).
• Deep experience with multimodal architectures and how encoders/projectors/LLMs fit together.
• Hands-on with RLHF/alignment for multimodal beyond text.
• Experience with video/audio sequence modeling and long-context processing.
• Track record of shipping models to production with latency optimization.
• Comfort with large-scale multimodal dataset curation and synthetic data generation.
• Familiar with MoE architectures and their tradeoffs for multimodal workloads.
• Strong engineering fundamentals: clean code, version control, testing, documentation

Responsabilités

• Train and fine‑tune large‑scale multimodal models from scratch or pretrained checkpoints.
• Extend models across modalities: vision-language, audio, video, speech.
• Design experiments: architecture changes, data mixes, training recipes.
• Build and maintain multimodal data pipelines from raw data to training-ready datasets.
• Train and optimize MoE architectures for efficient inference.
• Build alignment pipelines: SFT, DPO, reward modeling across modalities.
• Optimize models for production: quantization, distillation, batching, streaming.
• Deploy models end‑to‑end from research checkpoint to production serving.
• Define and monitor evaluation metrics and benchmarks for the product

Connaissances

Deep learning

PyTorch

Distributed training

Multimodal architectures

RLHF/alignment

Video/audio sequence modeling

MoE architectures

Model deployment

Data curation

Engineering fundamentals

Outils

DeepSpeed

FSDP

PyTorch tooling

Description du poste

White Circle is hiring a Multimodal ML Engineer in Paris. You will train and ship vision, audio, video, and speech models for an AI safety platform processing 100M+ API calls monthly.

You will extend models across modalities, design experiments, and build multimodal data pipelines for training-ready datasets, including synthetic data generation. You will deploy models end-to-end, define evaluation benchmarks, and optimize for production latency.