Researcher, Post Training

About the role

As a Researcher in Post Training at nyra labs, you will shape how speech models behave after pre‑training, improving accuracy, controllability, robustness and alignment with real user needs.

Key responsibilities

• Develop and test post‑training methods such as supervised fine‑tuning, preference optimization, distillation, feedback‑driven learning and reinforcement learning.
• Create high‑quality post‑training datasets via curation, annotation, synthetic generation and model‑assisted improvement.
• Improve model behavior for instruction following, transcription fidelity, uncertainty handling, multilingual performance and hallucination resistance.
• Build benchmarks, failure taxonomies and evaluation pipelines to assess genuine model improvements.
• Implement, debug and scale training and evaluation systems with a focus on reproducibility.
• Collaborate with research, product and clinical teams to adapt foundation models for specific speech and clinical use cases.
• Contribute to open publications, model releases, datasets and technical reports.

Required profile

• Practical expertise in post‑training techniques (fine‑tuning, RLHF, distillation, alignment, etc.).
• Strong deep‑learning background with proven PyTorch experience.
• Experience designing evaluations and diagnosing complex model behavior.
• Ability to write clean, production‑quality Python and debug distributed training systems.
• Record of publications, open‑source contributions or substantial independent research.
• Advanced degree (MSc, PhD) or equivalent practical experience in ML, speech processing, NLP or related fields.

Required skills

• Python
• PyTorch

What we offer

• Access to a uniquely large, therapist‑labeled neurological speech dataset.
• Opportunities to publish models, benchmarks and findings openly.
• Direct collaboration with founders, researchers, engineers and clinicians.
• Ownership over research direction from day one.
• Attractive compensation, phantom stock options and company benefits.
• Hybrid working model with a beautiful office in Vienna’s First District.