We’re hiring an Audio / Multimodal ML Engineer for a fast-growing AI infrastructure startup building the safety and control layer for large-scale AI systems.
AI-native product company operating in the AI safety & infrastructure space
Backed by international investors
Processing large-scale AI traffic across enterprise environments
Training and fine-tuning proprietary models for performance & reliability
Small, highly technical team shipping fast
The company builds the control and evaluation layer for AI systems - helping organizations define, test, and enforce how AI behaves in real-world environments.
Train and fine-tune large-scale audio & multimodal models
Design and run experiments (architecture, data mixtures, training strategies)
Build and optimize audio data pipelines
Improve inference speed, latency and production readiness
Deploy models end-to-end in low-latency environments
Define meaningful evaluation metrics beyond benchmark scores
Collaborate closely with research & engineering
This is a hands-on role where research meets production.
PyTorch-based training pipelines
Large-scale distributed training
Speech & audio modeling architectures
Multimodal model integration
Model optimization (quantization, distillation, streaming inference)
Production deployment & serving systems
(Deep technical stack shared during interviews)
3+ years training deep learning models in audio / speech domains
Strong experience with distributed training frameworks
Solid understanding of audio signal processing fundamentals
Experience shipping models to production (latency matters, not just metrics)
Experience building and maintaining data pipelines
Strong engineering hygiene (clean code, testing, versioning)
Experience with multimodal architectures
Experience with alignment / fine-tuning techniques
Experience working in AI infrastructure or model optimization
Competitive compensation + equity
Hybrid setup in Europe + relocation support
Comprehensive health coverage
Top-tier hardware & tools
Team off-sites
Budget for learning & AI tooling
connecthum