Staff Research Engineer (LLM Pre-Training)
Core
Developing Large Language Models from scratch for coding tasks and deploying them into production environments for JetBrains products.
Role type
Staff Research Engineer (LLM Pre-Training)
Builds
In-house foundation models for writing and coding assistance integrated into JetBrains products.
Domain
AI/ML, Large Language Models, Software Development Tools
Deliverable
production ML models
Required skills
LLM training from scratch, distributed training of multi-billion parameter models, NLP and transformer-based approaches, deep learning frameworks (PyTorch), production ML system design and deployment
Preferred skills
LLM inference frameworks (vLLM, DeepSpeed, TensorRT), LLM alignment techniques (RLHF/RLAIF), MLOps tools and practices, K8s and Kubeflow, scientific publications in NLP
Technologies
PyTorch, HuggingFace, Kubeflow, Weights & Biases, TeamCity, NVIDIA GPUs, Git
Responsibilities
Train LLMs from scratch on a large GPU cluster, collect and process pre-training and fine-tuning datasets, support and improve existing subsystems, convert business requirements into technical specifications
Seniority
Staff, hands-on IC with strategic impact