Multimodal ML Engineer

White Circle is an AI Safety company seeking a Multimodal ML Engineer to train and ship vision, audio, video, and speech models. You will work on large-scale multimodal models, alignment pipelines, and production optimization. The role involves building data pipelines and deploying models end-to-end.

• Train and fine-tune large-scale multimodal models from scratch; Extend models across modalities; Design and run experiments; Build and maintain multimodal data pipelines; Train and optimize MoE architectures; Build alignment pipelines (SFT, DPO, GRPO); Optimize models for production (quantization, distillation); Deploy models end-to-end; Define evaluation metrics.

• 3+ years training large-scale deep learning models in multimodal domains; Strong PyTorch with distributed training (DeepSpeed, FSDP); Deep experience with multimodal architectures; Hands-on with RLHF/alignment for multimodal; Experience with video/audio sequence modeling; Track record of shipping models to production; Comfortable with large-scale multimodal dataset curation; Familiar with MoE architectures; Strong engineering fundamentals.

• Understanding of audio signal processing fundamentals (spectrograms, mel features, noise reduction).