Remote AI Research Engineer: Agentic Post-Training

Alignerr Corp.; Córdoba; A distancia; ARS 181.747.000 - 363.493.000; Jornada completa

Descripción de la vacante

Tether is seeking an AI Research Engineer (Agentic Post-training) for a 100% remote worldwide role. You will drive innovation in post-training methodologies, focusing on agentic behaviors, tool use, and multi-modal models running from edge devices to cloud-scale systems.

Collaborate with research and engineering to push SOTA results, refine reasoning and coordination, and ensure robust tool invocation and self-correction across diverse deployments.

Formación

• Degree in Computer Science, ML, or related field; MS/PhD preferred.
• Experience with multimodal post-training workflows.
• Hands-on experience applying post-training at scale with distributed training.
• Proven track record of open-source contributions (GitHub, Hugging Face).
• Publications at AI conferences (NeurIPS, ICML, ICLR).

Responsabilidades

• Conduct end-to-end research and engineering initiatives to advance post-training of agentic and tool-use models.
• Drive cross-cutting model improvements including tool use, reasoning calibration, and multi-agent coordination.
• Design and enhance large-scale post-training systems, data pipelines, training workflows, and benchmarks.
• Develop rigorous evaluation suites and diagnostic tools for model readiness.
• Collaborate with cross-functional teams to determine production-ready integrations for major releases.

Conocimientos

Agentic AI

Tool use integration

Reasoning calibration

Multimodal reasoning

Educación

MS/PhD in CS/ML

Herramientas

Distributed training

GitHub

Hugging Face

Descripción del empleo

Tether is seeking an AI Research Engineer (Agentic Post-training) for a 100% remote worldwide role. You will drive innovation in post-training methodologies, focusing on agentic behaviors, tool use, and multi-modal models running from edge devices to cloud-scale systems.

Collaborate with research and engineering to push SOTA results, refine reasoning and coordination, and ensure robust tool invocation and self-correction across diverse deployments.