Frontier Research Engineer, Deep Learning Full Time

Frontier Research Engineer, Deep Learning | Foundational AI Company | Remote 

We are seeking innovative Research Scientists with a strong background in deep learning to tackle complex challenges in various applications. In this role, you’ll pioneer advancements by building foundational models that go beyond traditional tasks, focusing on capturing nuanced meanings and generating contextual responses.

Key Responsibilities:

• Design and implement experimental programs to create new AI models across diverse modalities and tasks.
• Execute large-scale training on extensive distributed computing infrastructures.
• Optimize model architectures for speed and memory efficiency, deploying new models for massive-scale production use.
• Clearly document and present findings and technical concepts for diverse audiences.
• Stay informed on the latest deep learning developments and their potential applications.

Ideal Candidate:

• You have a passion for AI and a desire to leverage data to solve challenging problems.
• You thrive on building systems from scratch and enjoy hands-on experimentation.
• You are data-driven, using iterative methods to address problems effectively.

Requirements:

• PhD in Physics, Electrical Engineering, Computer Science, or a related field.
• Experience designing and executing experimental programs to understand complex phenomena, with agility in adapting strategies as needed.
• Proven ability to build models from the ground up, encompassing data curation, architecture design, distributed training, and model optimization.
• Strong communication skills to simplify complex concepts for various audiences.
• Proficiency in software engineering, especially in Python and PyTorch.

Preferred Qualifications:

• Industry experience in developing deep learning models for intricate problems, with knowledge of various neural network architectures and loss mechanisms.
• Familiarity with advanced network architectures, particularly transformers.
• Understanding of parallelism paradigms for efficient distributed training.