Description

We are looking for a Middle ML Engineer/Researcher to join a research team working on the development of omnimodal solutions in the audio domain, as part of a large-scale project to create next-generation artificial intelligence systems.

Responsibilities

● Researching technologies for working with different audio modalities: speech, noise, music, sound effects

● Integrating audio, text, and visual modalities into a unified architecture

● Multimodal reasoning and stream synchronization (audio–text–vision)

● Researching and implementing state-of-the-art approaches (end-to-end, transformers, multimodal LLMs, diffusions)

Requirements

• Excellent Python 3, experience with PyTorch, bash, git, Docker, dvc, HF Transformers
• Good understanding of ASR, TTS, DSP ML, speech & audio processing
• Understanding of transformers, attention mechanisms, KV-cache, diffusion
• Skills in working with large audio datasets
• Understanding of MLOps practices: model monitoring, data drift, CI/CD

Will be a plus:

• Experience in speech, music domains, with voice assistants
• Experience with diffusion and autoregressive architectures for audio/music
• Experience with streaming / real-time systems
• Knowledge of multimodal LLM / VLM / Audio-LM
• Publications or research background in relevant fields

Conditions

• Comfortable modern office near Kutuzovskaya metro station
• Hybrid work format
• Annual salary review, annual bonus from 3 salaries
• Large gym and recreation areas
• Training system for professional and career development
• Extended voluntary health insurance policy from the first day of work and insurance for family
• Employee mortgage program with a discount of -1/3 from the current rate
• Free SberPrime+ subscription, discounts on products from partner companies
• Referral bonus for recommending friends to the Sber team.