About the job
About H:
At H, we are committed to redefining the potential of superintelligence through agentic AI. Our mission is to automate intricate, multi-step tasks traditionally carried out by humans, enabling AI agents to unlock the full spectrum of human capability.
We are on the lookout for the brightest minds in AI, those who are as passionate about developing safe and responsible technologies as they are about pioneering transformative agentic functionalities. We foster a collaborative culture where transparency, continuous learning, and contribution from every team member are valued.
About the Team: The Models team is at the forefront of creating the foundational models that fuel our cutting-edge agentic technologies. We concentrate on optimizing training techniques specifically designed for agent applications, ensuring peak performance with economical inference costs. Our scope includes the development of Large Language Models (LLMs) and Vision-Language Models (VLMs), empowering agents to interpret, comprehend, and interact within intricate environments. We actively enhance these models through innovative training strategies aimed at improving instruction adherence, tool utilization, and dynamic interaction using extensive reinforcement learning and reward modeling methods. Operating at the crossroads of research and practical application, we strive to convert pioneering research into actionable solutions that propel the evolution of AI. We invite enthusiastic and skilled individuals to join us in sculpting the future of superintelligent AI.
Key Responsibilities:
Design and train sophisticated LLMs and VLMs, including multimodal architectures.
Investigate and implement training methodologies to enhance capabilities such as instruction compliance and tool application.
Architect and refine data pipelines and training infrastructures for large-scale distributed training efforts.
Collaborate with multidisciplinary teams to integrate models into agentic AI frameworks.
Assess model performance and relay insights to stakeholders.
Stay abreast of advancements in LLMs, VLMs, and associated domains.
Requirements:
Technical skills:
Proficient programming skills (Python, Git).
In-depth knowledge of deep learning frameworks (PyTorch, JAX, TensorFlow).
Experience in large-scale distributed training of LLMs and VLMs.
Hands-on experience with LLM training and optimization processes.
