TLDR: Multimodal ML Engineer to train and ship vision, audio, video, and speech models for an AI safety platform that operates at 100M+ API calls/month.
About us
White Circle is an AI Safety company building the safety, reliability, and optimization layer for AI systems. At the core of our platform are policies – simple natural-language rules that define what an AI model should and shouldn’t do. We automatically test, enforce, and continuously improve these policies at scale.
We’ve raised $11M from top funds, founders, and senior leaders at OpenAI, Anthropic, HuggingFace, Mistral, DeepMind, Datadog, Sentry, and others
We process over 100M+ API calls every month
We fine-tune and train our own LLMs so they run faster and cheaper than any open or proprietary model
We’re a small, highly focused team. If you want to work deeply on hard problems, see your work ship to production quickly, and influence how AI safety is actually built – you’re the one we need.
You will
Train and fine-tune large-scale multimodal models (vision-language, audio, speech) from scratch and from pretrained checkpoints
Extend models across modalities: image understanding, video temporal modeling, long-context processing, and streaming audio
Design and run experiments: architecture changes, data mixes, training recipes
Build and maintain multimodal data pipelines — from raw images, video, and audio recordings to training-ready datasets, including synthetic data generation
Train and optimize MoE architectures for efficient multimodal inference
Build alignment pipelines: SFT, DPO, GRPO, reward modeling — across modalities, not just text
Optimize models for production: quantization, distillation, batching, streaming and low-latency serving
Deploy models end-to-end: from research checkpoint to production serving
Define evaluation metrics and benchmarks that actually matter for the product: visual QA, spatial reasoning, video comprehension, speech and audio understanding
You’ll fit right in if you
3+ years training large-scale deep learning models in multimodal domains (vision-language, audio, speech, or acoustic)
Strong PyTorch skills with hands-on distributed training experience (DeepSpeed, FSDP, or similar)
Deep experience with multimodal architectures — you understand how vision/audio encoders, projectors, and LLMs fit together (LLaVA, Qwen-VL, InternVL, Audio Flamingo, Omni Qwen, Audio Qwen, Whisper, HuBERT, Conformer, or similar)
Hands-on with RLHF/alignment for multimodal: GRPO, DPO, reward modeling — not just for text
Experience with video and/or audio sequence modeling: temporal modeling, long-context processing, efficient attention, streaming inference
Track record of shipping models to production: you've hit latency targets and optimized inference, not just reported benchmark scores
Comfortable with large-scale multimodal dataset curation: image-text pairs, video-instruction data, audio preprocessing, augmentation, synthetic data generation
Familiar with MoE architectures and their tradeoffs for multimodal workloads
Strong engineering fundamentals: clean code, version control, testing, documentation
A big plus:
Understanding of audio signal processing fundamentals (spectrograms, mel features, noise reduction)
Why White Circle
Paid time off in line with your local regulations, no matter where you work from
Work from Paris (hybrid) with a relocation package available, or work from London (note: we are unable to provide relocation support for London-based roles)
Comprehensive medical insurance for our France-based team (please note that we are in the process of setting up our UK office and therefore cannot offer medical insurance for London-based roles yet)
Meaningful equity package
All the hardware, tools, and services you need
Covered subscriptions for AI agents and IDEs
Team off-sites twice a year: we’ve recently been to the Alps and to Saint-Tropez
How we hire
Introductory call with HR (25 min)
Take-home test task
Technical interview with Head of Applied Research (60 min)
Final conversation with our CEO (45 min)
