Сбер

Open 125d RemoteCISC-level

2 months ago

Руководитель направления Online RL (STEM) / Post-Training LLM в команду GigaChat

Distributed Training Fine Tuning LLM Python PyTorch +1 skills

Lead the online reinforcement learning team for STEM-focused LLM post-training at GigaChat, designing RLHF pipelines, reward models, and data curation to improve reasoning in math, physics, and other sciences.

650K – 850K RUB

2 months ago

Online RL (General) / Post-Training LLM в команду GigaChat

Distributed Training Fine Tuning LLM Python PyTorch +1 skills

Lead the online reinforcement-learning team for GigaChat, designing post-training methods, reward models, and data pipelines to make the LLM more helpful and user-friendly.

650K – 850K RUB

2 months ago

NLP Engineer (GigaChat Pretrain)

AI ChatGPT CUDA LLM NLP +3 skills

Build and optimize large-scale generative language models (LLMs) for GigaChat, focusing on Russian-language performance, distributed training, and efficiency improvements.

NLP-инженер в GigaChat Pro

AI Fine Tuning GPU LLM Machine Learning +5 skills

NLP engineer working on GigaChat Pro, a bank employee copilot assistant that interacts with clients. Daily tasks include developing LLM/BERT-based models, optimizing pipelines, and testing new AI-driven business solutions to enhance the assistant’s functionality and performance.

Open 125d RemoteCISSenior

Senior Research Engineer (LLM Pretraining)

Senior Research Engineer focused on large language model pretraining, optimizing training recipes, architectures, and distributed training infrastructure using PyTorch on H100/B200 clusters.

AI CI/CD PyTorch

Open 114d RemoteCISC-level

Руководитель направления ML Pretrain LLM

AI Distributed Training LLM Machine Learning Python +1 skills

Lead the architecture and pretraining of large language models for GigaChat, designing MoE and Transformer-based models, scaling laws, and long-context features on H100/B200 clusters.

Open 114d RemoteCISC-level

MLE Руководитель направления RL Efficiency (Lead)

CUDA Distributed Training GPU Kubernetes LLM +7 skills

Lead a team to optimize and scale the reinforcement learning pipeline for GigaChat, focusing on speed, stability, and cost-efficiency of online RL/RLHF/RLVR training and inference.

Open 125d RemoteCISSenior

Senior ML Engineer с опытом в обучении генеративных моделей: TTS, LLM, CV

AI LLM Machine Learning RLHF

Senior ML Engineer builds and trains generative AI models for text-to-speech, voice cloning, and multimodal systems using LLM, TTS, and RL techniques.

Open 125d RemoteCISC-level

Руководитель аналитики метрик качества GigaChat (LLM Evaluation)

AI BigQuery CI/CD ClickHouse Grafana +10 skills

Leads the quality-metrics system for GigaChat, a large language model, designing benchmarks, analyzing logs, and embedding metrics into CI/CD to catch regressions and guide model improvements.

NLP-инженер (GigaChat)

ChatGPT C++ CUDA LLM NLP +2 skills

Develops and improves a Russian generative AI model (GigaChat) by training large language models, optimizing distributed training frameworks, and enhancing multimodal capabilities using Python, PyTorch, C++, and CUDA.

ML-инженер (AI Quality / Evaluation) в команду GigaChat

AI Bash Bootstrap CI/CD Git +5 skills

Мы развиваем GigaChat и ищем сильного ML-инженера в команду AI Quality. Наша команда — это главный навигатор развития модели. Мы не просто тестируем сборки, мы исследуем новые возможности GigaChat, сопоставляем его…