Post-training - Model Fusion Research Engineer

Our team is responsible for the post-training phase of ChatGPT, transforming large pre-trained models into powerful, safe, and user-friendly chatbots. We collaborate across the company to enhance ChatGPT's safety, speed, intelligence, utility, and overall capabilities. We integrate these improvements into the final models powering our production ChatGPT and API services, impacting millions of users worldwide.

We are seeking an engineer to accelerate the deployment of improvements to our models. You will collaborate with diverse teams handling various facets of the system, including core capabilities, multimodal integration (speech, images, and video), tools, and more. This role offers a unique opportunity to shape the future of ChatGPT, working across the technology stack, from optimizing low-level components like GPU kernels and network traffic to mastering the intricacies of RL post-training.

The ideal candidate has a robust technical background in areas such as data technologies, reliable software engineering, production ML model development, and cross-functional collaboration. While research experience is not required, a deep understanding of ML fundamentals and large-scale deep learning is essential for troubleshooting and analyzing complex system and ML issues. Excellent verbal and written communication skills, along with strong project management abilities, are crucial as you will collaborate with both technical research teams and non-technical product teams across the company.

Key Responsibilities:

• Lead the regular release process of ChatGPT and API models
• Dive deep into large ML codebases to diagnose and resolve systems and ML issues
• Ensure a stable, smooth, and user-friendly model development process
• Collaborate with cross-functional teams to design new features and integrate enhancements into ChatGPT

Sample projects include:

• Consulting with product teams on how to best improve ChatGPT
• Integrating new multimodal features or tools into ChatGPT
• Working with data scientists to understand the impact of new models on users

This role is based in San Francisco, CA, with a hybrid work model of 3 days in the office per week. Relocation assistance is offered to new employees.