Remote AI Research Engineer — Agentic Post-Training

Impulso Media; Tandil; A distancia; ARS 216.357.000 - 278.173.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Remote worldwide

Descripción de la vacante

Tether is seeking an AI Research Engineer (Agentic Post-training) to advance post-training methods for agentic AI across edge devices, with focus on tool use and multi-modal systems. You will craft data pipelines, evaluation suites, and research-driven implementations, collaborating across global teams.

The role emphasizes leadership in post-training algorithms, sustained experimental rigor, and open-source contributions to push SOTA capabilities in agentic workflows.

Formación

• Deep expertise in large language model architectures and post-training for agentic workflows.
• Experience with tool use fine tuning, function calling, and RLHF for multi-turn tasks.
• Proven track record in open-source contributions related to agentic systems.

Responsabilidades

• Conduct end-to-end research and engineering to advance post-training of agentic models.
• Design data pipelines, training workflows, evaluation frameworks, and benchmarks.
• Collaborate with product and engineering teams to deploy agentic capabilities.
• Curate training data with tool use trajectories and reasoning chains.
• Publish results in top AI conferences and contribute to open-source projects.

Conocimientos

Agentic AI

Tool use

Multimodal reasoning

Distributed training

Research & development

Educación

MS/PhD preferred

Herramientas

GitHub

Hugging Face

Descripción del empleo

Tether is seeking an AI Research Engineer (Agentic Post-training) to advance post-training methods for agentic AI across edge devices, with focus on tool use and multi-modal systems. You will craft data pipelines, evaluation suites, and research-driven implementations, collaborating across global teams.

The role emphasizes leadership in post-training algorithms, sustained experimental rigor, and open-source contributions to push SOTA capabilities in agentic workflows.