LLM Researcher

About the Role We are looking for an experienced LLM Researcher to help us build intelligent, automated systems that enhance customer-facing applications. In this role, you are expected to utilize, optimize and finetune existing language models or train new ones, more adapted to our use cases - and turn successful prototypes into reliable, maintainable, and observable production services. Your work will directly contribute to the development of dynamic, real-time commentary content for games such as FIFA, NBA, or interactive avatars. This is a remote role with an option to join our Prague office, offering the opportunity to work with cutting-edge LLM frameworks, retrieval-augmented generation (RAG), and multimodal models that process both language and vision. You will collaborate with a small, fast-moving team of ML engineers, product developers, and domain experts to design systems that combine reasoning, tool use, and creative content generation. Responsibilities Must-have: - Solid experience with LLMs in production environments, including training, fine-tuning, inference, and tool integration. - Experience building RAG-based systems and working with vector or graph databases. - Experience deploying, monitoring and integrating these systems into the overall solution. - Strong understanding of function calling, structured output generation, and agentic reasoning workflows. - Proficiency in Python and key libraries (e.g., PyTorch, Hugging Face Transformers, FastAPI). Nice-to-have: - Experience with sports broadcasting or data-driven content generation (e.g., live commentary, analytics). - Familiarity with multimodal LLMs and integrating textual and visual inputs. - Familiarity with tools like vLLM, Triton, or DeepSpeed for efficient model serving. - Contributions to open-source AI tools or research publications in NLP, multimodal AI, or agent systems.