Founding Speech Model Performance Engineer

• We pay top of the market → We want serious talent in the team, and we benchmark compensation accordingly.

• Equity → All full-time, permanent roles include stock options - we want everyone to share in the upside as we build.

• Flexible benefits: Health insurance, gym, and more via Cobee.

• L&D: Annual budget (up to €1000) for training, courses, or conferences.

• Remote work support: Monthly stipend (up to €50) to cover wifi or other costs.

• Great equipment: Laptop of your choice plus €500 to set up your workstation in year one, with top-ups in the following years.

• Time off: Additional +5 days vacation on top of the statutory minimum

About The Role → You'll make speech models fast.

• Strong background in ASR/TTS.

• PyTorch/ONNX experience.

• Familiar with GPU profiling and optimisation.

• Fluency in English.

SLNG is building the backbone for real-time speech AI, enabling developers to run voice applications anywhere in the world with local compliance and ultra-low latency.

Our founding team comes from the core of AI and developer tooling in the USA, with experience scaling platforms trusted by the world's best builders. Meet Luke, Founder & CEO, and Ismael, Founder & CPO at SLNG.

We're bringing the San Francisco mindset to Europe with our first hub in Barcelona, and we have the support of leading international VCs and Angel investors who are backing the SLNG journey.

Challenge → Today, most speech infrastructure is US-centric, unreliable, and difficult to deploy globally. SLNG is changing that by delivering a platform that is:

• Local: deployed close to users and compliant with regional regulations

• Fast: designed for real-time voice experiences

• Open: built to integrate with the tools and workflows developers already use.

From real-time transcription to voice AI applications, SLNG is creating the Voice AI gateway that will empower the next generation of speech-powered products.

As Speech Model Performance Engineer, you'll work closely with Ismael and the founding tech team to shape the technical foundation of SLNG. From TTS voices to multilingual ASR, you'll benchmark, optimise, and productionize speech inference at scale.

• Quantise neural TTS and STT models to run with minimal latency on heterogeneous GPU hardware.

• Benchmark ASR models across dialectal variations, measuring Word Error Rate (WER) and latency trade-offs.

• Implement continuous batching and KV-cache reuse for streaming inference.

• Profile GPU utilisation with CUDA kernels to identify bottlenecks in large-scale inference.

• Hybrid by design → 3 days/week in the Barcelona office for collaboration and culture.