Remote AI Research Engineer - Agentic Post-Training Expert

Mindrift; La Pampa; A distancia; ARS 139.086.000 - 247.265.000; Jornada completa

Descripción de la vacante

Tether seeks an AI Research Engineer focusing on agentic post-training, tool use, and multi‑turn reasoning. You will refine pre‑trained models to enable autonomous tool invocation and multi-step decision making across edge devices and cloud environments.

Work on multimodal systems, data pipelines, and distributed training, collaborating with cross‑functional teams to push SOTA capabilities and publish findings in top AI venues.

Formación

• Deep expertise in large language model architectures and post-training.
• Experience with agentic workflows, tool use, and multi-turn reasoning.
• Strong publication record in AI conferences.

Responsabilidades

• Conduct end-to-end research and engineering for post-training of agentic and tool-use models.
• Drive model improvements: factuality, tool use, multi-agent coordination, and reasoning calibration.
• Design scalable post-training pipelines, training workflows, evaluation frameworks, and benchmarks.
• Develop rigorous evaluation suites and diagnostic tools for deployment readiness.
• Collaborate with tooling, product, and training teams to enhance agentic capabilities.
• Identify production-ready integrations for major model releases.

Conocimientos

LLM architectures

Agentic post-training

Tool use fine-tuning

Reinforcement learning

Multimodal systems

Distributed training

Open-source contributions

Academic publications

Educación

Degree in Computer Science or ML (MS/PhD preferred)

Herramientas

GitHub

Hugging Face

Descripción del empleo

Tether seeks an AI Research Engineer focusing on agentic post-training, tool use, and multi‑turn reasoning. You will refine pre‑trained models to enable autonomous tool invocation and multi-step decision making across edge devices and cloud environments.

Work on multimodal systems, data pipelines, and distributed training, collaborating with cross‑functional teams to push SOTA capabilities and publish findings in top AI venues.