MWS AI is a part of the MTS Web Services ecosystem where future AI solutions are created. Voice assistants and chat bots can answer any question of the client, and computer vision - recognize faces and emotions. We brought together strong developers, the most powerful supercomputer, and breakthrough ideas to make people’s lives more comfortable and safe and help business.
We are seeking a Researcher to drive the development and training of Large Language Models (LLMs), Vision-Language Models (VLMs), and omni-modal models. to help develop and train Large Language Models (LLMs), Vision-Language Models (VLMs), and omni-modal models. The successful candidate will work at the intersection of research and practical systems, contributing to model design, experimentation, and scalable training pipelines under the guidance of more senior researchers and engineers.
This role is ideal for candidates who already have some hands-on experience with deep learning and are motivated to grow into large-scale LLM/VLM research and engineering.
Key Responsibilities: - Contribute to the design, training, and evaluation of LLMs, VLMs, and omni-modal models
- Run experiments, analyze results, and help iterate on model architectures and training recipes
- Assist in developing and optimizing training and inference pipelines for latency, throughput, and cost efficiency
- Help implement and maintain distributed training workflows on multi-GPU and (over time) multi-node clusters
- Collaborate with engineers and product teams to transfer research outcomes into real-world applications and products
- Contribute to writing internal reports, technical documentation, and, when appropriate, research papers
- Stay current with the latest advances in machine learning, deep learning, and large-scale model training
Required Qualifications: - Bachelor’s degree (completed or in final year) or Master’s degree in Computer Science, Electrical Engineering, Mathematics, or a related field, or equivalent practical experience
- Good oral and written English at a minimum level of B1-В2
- Solid background in mathematics, including linear algebra, probability, statistics, and optimization
- Strong programming skills in Python, with experience in deep learning frameworks such as PyTorch (preferably), TensorFlow, or JAX
- Practical experience training deep neural networks (e.g., for NLP, computer vision, recommendation, or related tasks), including data preprocessing, model implementation, and debugging
- Familiarity with core machine learning concepts (overfitting, regularization, evaluation metrics, cross-validation, etc.)
- Ability to read and implement methods from recent ML research papers
- Ability to work both independently and collaboratively in a fast-paced research and engineering environment
- Good communication skills, including the ability to clearly present experimental results and technical ideas
Beneficial Qualifications: - Hands-on experience with training or fine-tuning LLMs, VLMs, or other large foundation models (e.g., using HuggingFace, LLaMA Factory, or similar frameworks)
- Familiarity with distributed and parallel training concepts (e.g., data/model parallelism, mixed-precision training, gradient accumulation, checkpoints)
- Experience using job schedulers or orchestration tools (e.g., SLURM, Kubernetes, Ray) in any setting (uni-versity, research lab, or industry)
- Exposure to model efficiency techniques such as quantization, pruning, distillation, low-rank methods, or sparsity
- Experience contributing to open-source ML projects or maintaining reproducible ML codebases
- Experience building and deploying smaller-scale ML models into real products, services, or internal tools
Extra Beneficial Qualifications:
- Ph.D. in Computer Science, Electrical Engineering, Mathematics, or a related field with a focus on machine learning or deep learning
- Prior research or engineering experience specifically focused on large-scale LLMs, VLMs, or omni-modal models (pre-training, alignment, evaluation, or compression)
- Deep hands-on experience with distributed training on multi-GPU or multi-node clusters (e.g., FSDP, Deep-Speed, ZeRO, pipeline or tensor parallelism)
- Strong track record of work on model compression, acceleration, or efficient inference for large models (e.g., production-grade quantization, sparsity, distillation, or low-rank adaptation)
- Publications at top-tier conferences such as NeurIPS, ICML, ICLR, CVPR, ICCV, ACL, EMNLP, or similar venues, or strong evidence of research impact (e.g., widely used open-source projects)
- Experience leading ML projects end-to-end, from problem definition and data collection to deployment and monitoring
What will you get from us:
- 5/2, partial or full remote working scheduling, flexible start of the day
- Fixed monthly salary + quarterly bonuses
- Comfortable office 5 minutes walk from the metro station (Moscow)
- Medical insurance with dentistry, travel insurance, life insurance, discounts on car insurance, etc.
- Participation in conferences and meetups, trainings at the expense of the company
- Free mobile communications
- Annual paid leave - 31 days
- Special offers and discounts from MTS partners
