Research Manager

Revoize Inc. is a deep-tech company building real-time, on-device speech-to-speech translation (S2ST) that preserves the speaker's voice and emotion. The company is hiring proven researchers to build a research team and reach on-device S2ST models within the next 12 months.

This is the first senior research hire. The Research Manager will own the research roadmap, lead and mentor the research team (a codec researcher today, plus the Research Engineer hired in parallel) and stay hands-on in key modelling decisions. Reports directly to the CEO.

Key information:

Salary:

• 8 000 - 13 500 USD net/month - pure B2B contract

Wymagania

It is vital that you have:

• PhD in ML / speech / audio, or an equivalent research track record.

• 8+ years in ML (PhD years count), 5+ in speech/audio, of which 2+ with modern generative models (neural codecs, speech LMs, diffusion);

• hands-on, end-to-end training of large generative speech models (speech LMs, neural audio codecs, ASR / TTS / speech translation), not only fine-tuning;

• 2+ years leading researchers or a small research team: roadmap, prioritisation, mentoring, hiring;

• publications at top venues (e.g. Interspeech, ICASSP, NeurIPS, ICLR, ACL);

• practical sense for real-time constraints: latency, model size, on-device deployment;

• independent, hands-on, startup mindset;

• fluent English (min. C1).

We will be more than happy if you have:

• Model compression (quantization, distillation);

• on-device inference (ONNX, Core ML, TFLite, NPU toolchains);

• streaming / simultaneous translation;

• reinforcement learning;

• DSP;

• LLM training;

• cloud (Azure, AWS);

• multilingual work (Polish a plus).

Technology stack:

• Python, PyTorch;

• generative speech / audio models (speech LMs, neural audio codecs);

• ASR, TTS, machine translation;

• multi-GPU distributed training;

• TensorBoard;

• Linux, Git, Docker.

Obowiązki

You will be responsible for supporting our team in:

• own the research roadmap for speech-to-speech translation and the speech codec; turn company milestones into experiments and deliverables,

• lead and mentor the research team; set standards for reproducible experiments and reviews; support hiring,

• stay hands-on in key architecture and training decisions;

• define evaluation (translation quality, latency, speaker and emotion preservation) and make go/no-go calls on models;

• work with the data lead and MLOps on training data and infrastructure;

• drive models towards real-time, on-device deployment.

Oferujemy

What do we offer:

• work 100% remote;

• b2b contract;

• in addition to the cash salary, Revoize offers an ESOP (employee stock option plan);

• Claude Max – Extended Use (Max 5x) or Claude Max – Max Flexibility (Max 20x).