Software Engineer - Multimodal

The multimodal team at xAI creates AI experiences beyond text, enabling understanding and generation of content across various modalities including image, video, and audio. The role involves developing and optimizing data pipelines related to multimodal data such as acquisition, preprocessing, data loading, visualization, and management. Responsibilities include building tools for multimedia data acquisition, petabyte-scale high-throughput data processing systems, low-latency data decoding and loading pipelines for large-scale training of multimodal models, and visualization and management tools for datasets. Ideal candidates are experts in software for large-scale distributed machine learning systems, Spark, GPUs, Kubernetes, and JAX or PyTorch, and follow best software engineering practices including CI/CD, code quality, testing, and performance. The tech stack includes Python, JAX, Rust, Spark, and CUDA. The interview process includes an initial phone interview followed by four technical interviews and a presentation of past work and vision with xAI. All interviews are conducted via Google Meet.