This role focuses on critical post-training and reinforcement learning challenges for AI models, including reward modeling, preference optimization (RLHF/DPO), and RL for improving reasoning, truthfulness, and real-world capabilities. The goal is to build useful models through these techniques.

xAI is seeking a Member of Technical Staff for their Multimodal Understanding team to advance superhuman multimodal intelligence. The role involves working across the full stack of multimodal AI, from data curation and pre-training to post-training, inference, evaluation, and end-to-end product experiences. Responsibilities include designing and optimizing large-scale distributed systems, developing data pipelines, advancing multimodal capabilities, creating evaluation frameworks, and innovating on algorithms and scaling paradigms. The role requires hands-on experience with multimodal pre-training/post-training/fine-tuning, proficiency in Python and ML frameworks, and proven track record in building large-scale distributed ML systems and data pipelines.

The role focuses on building and improving voice AI models for natural, low-latency spoken interactions. This involves large-scale data curation, speech-language model pre-training and post-training, and developing a comprehensive evaluation framework. The goal is to integrate these models into real-time applications for a global scale deployment.

xAI is seeking a multimodal engineer to develop AI experiences beyond text, focusing on image and video understanding and generation, with audio integration. Responsibilities include data curation, modeling, training, inference serving, and product integration, covering pretraining and post-training phases. The role involves creating engineering agendas, improving data quality, designing evaluation frameworks, implementing efficient algorithms for real-time inference and serving, and developing scalable data pipelines for multimodal datasets. Collaboration with product teams for production integration and iteration based on user feedback is key.

This role focuses on improving the epistemic rigor and truthfulness of AI models, specifically ensuring they reason carefully, avoid motivated reasoning, and communicate uncertainty appropriately. Responsibilities include assessing model outputs for accuracy and logical coherence, identifying fallacies, writing exemplary reasoning, and constructing adversarial examples. The role requires a strong analytical background, a track record in forecasting or rigorous analysis, and deep knowledge in fields like philosophy of science, cognitive psychology, or statistics.

This role focuses on training and refining Grok (an LLM) to enforce terms of service, minimize risks, and prevent harmful content. Responsibilities include monitoring content, processing appeals, auditing automations, labeling data, and collaborating to improve AI defenses. Requires expertise in improving LLMs for safety and enforcement, experience in online safety, policy interpretation, data analysis, and ethical reasoning. Preferred experience includes Trust and Safety roles in social media and red-teaming LLMs.

Manager for Safety Operations at xAI, responsible for leading a team that trains and refines Grok (an LLM) to enforce terms of service, minimize risks, and prevent harmful content. The role involves managing analysts, overseeing data labeling, ensuring quality curated data for ethical alignment, identifying abuse vectors, and improving AI defenses. Requires leadership experience in AI-driven operations and expertise in LLM improvement for safety and efficiency.

This role involves contributing to AI model training by curating code examples, providing solutions, and making corrections in specialized programming languages. The candidate will evaluate and refine AI-generated code for efficiency and scalability, and collaborate with teams to enhance AI-driven coding solutions to meet enterprise-level quality and performance benchmarks.

This role focuses on improving the social cognition and emotional intelligence of an AI model (Grok) by teaching it to interpret emotional subtext, social context, and user intent, and respond with appropriate calibration. Responsibilities include detecting and interpreting emotional states, teaching models specific response styles, correcting tone-deaf outputs, building datasets to immunize the model against social tactics, and ensuring cross-cultural sensitivity. The role requires direct clinical experience and expertise in psychological frameworks.

This role focuses on refining the stylistic output of an AI model, specifically Grok, by ranking and rewriting responses for elegance, voice consistency, and aesthetic impact. The tutor will curate training data from high-quality writing and teach the model nuances of language use.

This role focuses on defining and safeguarding the personality and voice of an AI model (Grok), involving reviewing and scoring model responses, writing/revising sample responses, maintaining personality consistency, creating labeled training datasets for specific conversational elements, and developing evaluation tasks with engineering teams. The goal is to enhance user engagement and entertainment while ensuring factual accuracy and a distinctive, witty, culturally fluent voice.