Machine Learning Researcher

nyra health; Wien; On-site; EUR 70,000 - 110,000; Full time

Job summary

nyra health is seeking an ML Researcher to own research on neurological speech data, taking ideas from first experiment through to models that run in myReha and nyra insights. You will work closely with our Engineering team to bring models into production on-device and in the cloud.

You’ll collaborate with therapists, clinicians and product teams to decide which problems are worth solving and shape nyra’s ML research roadmap with founders and academic partners on publications and joint research

Qualifications

• MSc or PhD in ML, NLP, speech processing or related field.
• Strong background in speech/audio processing and related architectures.
• Proficient in PyTorch and scalable model training techniques.
• Proven track record of research publications or open-source contributions.
• Ability to write production-quality Python code and collaborate with engineers.

Responsibilities

• Develop and improve models for neurological speech recognition and pronunciation scoring.
• Prototype models for therapy content generation and clinical decision support.
• Explore multimodal data combinations to enhance patient recovery modeling.
• Design evaluation frameworks and training pipelines for clinical contexts.
• Collaborate with founders, clinicians, and academic partners on research roadmap.

Skills

MSc/PhD in ML or NLP

Speech processing

PyTorch

Model training at scale

Publication track record

Education

MSc or PhD in ML/NLP

Tools

Whisper

wav2vec

HuBERT

Claude Pro

Job description

Every morning, thousands of people relearn to speak with myReha, and your models listen to every attempt.

nyra health builds the digital tools that make neurological rehabilitation work inside and outside clinic walls: myReha, the patient therapy app, and nyra insights, the clinician platform. 100+ clinics, 100K+ patients across DACH. 3× MRR growth target for 2026. €20M Series A. Team of 30+, scaling to ~50 this year. Ownership, excellence and patients first — that’s how we drive progress together.

As our ML Researcher, you’ll report to our Head of ML and own research on nyra health’s neurological speech data, taking ideas from first experiment through to models that run in myReha and nyra insights. You’ll work closely with our Engineering team to bring models into production on-device and in the cloud. You’ll also work alongside our therapists, clinicians and product team, who know the patient problems firsthand and will help you decide which ones are worth solving. You’ll collaborate directly with the founders on nyra’s ML research roadmap, and with our academic partners on publications and joint research projects. It’s a role with a lot of autonomy and real scientific ownership. Your work goes straight to patients relearning how to speak, not to a benchmark leaderboard.

What you'll shape

• Speech and language models: Develop and improve models for neurological speech recognition, pronunciation scoring, and phoneme-level analysis — building on our proprietary models

• Clinical AI: Research and prototype models for therapy content generation, patient progress prediction, and clinical decision support

• Multimodal approaches: Explore combinations of speech, text, and behavioral signals to build richer models of patient recovery

• Data pipeline and evaluation: Design evaluation frameworks, build training pipelines, and establish rigorous benchmarks for model performance in clinical contexts

• Publication and collaboration: Contribute to scientific publications and collaborate with academic partners on research projects

• Production integration: Work with Engineering to deploy models into our production stack — ensuring they run reliably at scale on-device and in the cloud

• Research direction: Help shape nyra's ML research roadmap, identifying high-impact opportunities at the intersection of AI and neurological rehabilitation

• AI-native workflow: You use AI tools daily — Claude Code, Cursor, Copilot, or similar — not as a novelty, but as your default way of working. You write better code faster because of it.

• Multi-agent fluency: You're comfortable orchestrating multiple AI agents, chaining prompts, and building workflows where AI handles the heavy lifting while you steer the outcome.

• Tooling mindset: You actively shape your own dev environment — custom scripts, automation, MCP servers, CLI tools — and you're always looking for ways to remove friction from your workflow.

• ML research experience: MSc or PhD (or equivalent practical experience) in machine learning, speech processing, NLP, or a related field

• Speech and audio expertise: Strong background in speech recognition, speech synthesis, audio processing, or related areas (experience with Whisper, wav2vec, HuBERT, or similar architectures)

• Deep learning proficiency: Hands-on experience with PyTorch, model training at scale, and modern training techniques (LoRA, RLHF, distillation, etc.)

• Research rigor: Track record of publications, open-source contributions, or documented research projects

• Engineering skills: Ability to write clean, production-quality Python code — not just notebooks

• Evaluation mindset: Experience designing experiments, defining metrics, and drawing valid conclusions from data

Beyond your CV

• Scientifically rigorous: You care about doing things right — proper evaluation, reproducibility, and honest assessment of what works and what doesn't

• Impact-oriented: You're motivated by building models that ship and make a difference, not just benchmarks

• Curious and collaborative: You enjoy working with clinicians, therapists, and product teams to understand the problems worth solving

• Self-directed: You can identify promising research directions, design experiments, and drive projects forward with minimal guidance

• Passionate about speech and language: You find the intersection of AI and human communication fascinating — especially in populations underserved by existing technology

Why nyra health

• Where your research lands: A better pronunciation score means more accurate feedback for a patient relearning to speak, every single day.

• Patients first, always: We build for better patient outcomes. Your work gives our teams more time to focus on the people delivering and receiving rehabilitation.

• Ownership & direct access: See the impact of your work and take responsibility for outcomes. Founders, ML engineers and practicing clinicians work closely together, so decisions move fast and you’re part of them.

• Transparent, direct culture: Open communication and honest feedback are how we operate.

• AI-native: Every team member has Claude Pro and is expected to use AI tools daily.

• Some of the best Mario Kart drivers in the Austrian startup scene

The Process

• Intro call (~30 mins): Background, expectations, and an overview of nyra health and the role.

• Technical deep-dive: A research discussion covering your past work, approach to a relevant problem, and a technical scenario, with our ML and Engineering leads.

• Meet with Founders: Discuss your research vision, how you'd leverage nyra's data assets, and how you see the future of AI in neurological care.