Push the state of the art in voice AI. Train, evaluate, and deploy the speech and language models that power Orova's agents in production.
About the role
We are looking for an ML / NL engineer to push the state of the art in voice AI. You will own the model side of conversational agents — from training data curation through fine-tuning, evaluation, and production deployment.
This role sits between research and engineering. We expect you to read papers, run experiments quickly, ship to real customers, and instrument the results. If you want your model improvements to land in front of millions of conversations within a sprint, you'll feel at home here.
Key responsibilities
Train and fine-tune large language and speech models for production voice agents.
Build evaluation pipelines that catch regressions on safety, factuality, and latency before release.
Run experiments with real call data to improve handling of accents, code-switching, and noisy environments.
Collaborate with platform engineering on inference optimization, quantization, and caching.
Stay close to research, prototype quickly, and decide what makes it into the product.
Necessary skills
Strong background in machine learning, NLP, or speech research with publications or production work.
Experience training and fine-tuning transformer-based models (LLMs, ASR, TTS).
Proficiency with PyTorch or JAX, GPU-accelerated training, and distributed training frameworks.
Familiarity with retrieval-augmented generation, prompt engineering, and evaluation harnesses.
A track record of moving research into reliable production systems.
Apply for this role
We read every application within five working days.