Machine Learning Developer for LLM Team

About the role

Yandex Kyrgyzstan is looking for an ML Developer to join the LLM team that builds language models for the Alice voice assistant. The role is hybrid and covers the full production cycle, from data preparation and model training to evaluation and deployment.

Key responsibilities

• Fine‑tune language models using SFT, DPO or Reinforcement Learning.
• Configure system prompts and shape model behavior.
• Generate and improve multi‑step conversation corpora, filter examples with LLMs.
• Compare automated scoring (LLM‑as‑a‑Judge) with expert annotations.
• Teach the model to use correct functions, ask clarifying questions and correctly interpret numbers, dates, names and locations.
• Develop benchmarks for conversations and commands.
• Ensure overall model quality – authenticity, context memory, personality and AI safety.
• Integrate models into product infrastructure, optimise response times and inference.
• Contribute to the development process from closed beta through post‑launch.

Required profile

• Strong Python knowledge and experience building ML/NLP systems with PyTorch.
• Solid understanding of how transformers and large language models work.
• Hands‑on experience fine‑tuning LLMs using SFT, DPO or RL (production, research or pet projects).
• Experience with QA processes: creating test datasets, analysing data slices, using LLM judges and running A/B tests.
• Independent mindset, ability to grasp complex systems and drive results while collaborating with editors and adjacent teams.

Required skills

• Python
• PyTorch
• Transformers
• LLM fine‑tuning (SFT, DPO, RL)
• QA testing and A/B testing
• vLLM and distributed training

What we offer

• Psychological support in multiple languages.
• Life and private health coverage (subject to location‑specific conditions).
• Life and private health insurance (different conditions apply based on hiring location and position type).