Applied Frontier AI Researcher: Multimodal Agents

Prime Intellect, Inc.; United States; À distance; USD 150 000 - 300 000; Plein temps

Avantages offerts par ce poste

Cash compensation + equity

Flexible work (SF, NY, London, or 100%

Visa sponsorship & relocation

Professional development budget

Team off-sites & conference attendance

Résumé du poste

Prime Intellect, Inc. is building the open superintelligence stack and a full-stack platform for post-training at frontier scale.

You will help turn real workflows into multimodal environments, reward signals, and evaluation systems across browsers, desktops, and cross-application workflows, with externally facing research duties and customer collaboration. The role focuses on post-training and evaluation, ensuring recovery from interface changes, and advancing production models with a strong

Qualifications

• Relevant experience in multimodal agents, computer-use interfaces, or long-horizon agent evaluation.
• Hands-on experience building and debugging model-training loops and evaluations.
• Strong experimental judgment with clear failure attribution and adaptability.

Responsabilités

• Turn workflows into tasks, environments, evaluations, and reward signals.
• Run post-training experiments against baselines and deploy effective results.
• Collaborate with customers and researchers to advance open frontier AI capabilities.
• Develop environments and agents across browser, desktop, and cross-application contexts.

Connaissances

Multimodal agents

Browser automation

RL

Imitation learning

HCI

Embodied agents

Long-horizon eval

Description du poste

Prime Intellect, Inc. is building the open superintelligence stack and a full-stack platform for post-training at frontier scale.

You will help turn real workflows into multimodal environments, reward signals, and evaluation systems across browsers, desktops, and cross-application workflows, with externally facing research duties and customer collaboration. The role focuses on post-training and evaluation, ensuring recovery from interface changes, and advancing production models with a strong