AI Researcher – Agentic & Multimodal Systems

About the role

We are seeking an AI Researcher to join a global AI research organization. The role focuses on agentic AI, building next‑generation systems that learn from large‑scale real‑world interaction data across text, audio, logs, and structured workflows.

Key responsibilities

• Conduct advanced research on agentic AI systems, improving reasoning, planning, and tool use.
• Design and experiment with learning frameworks such as Retrieval‑Augmented Generation (RAG), fine‑tuning, RLHF, DPO, and GRPO.
• Develop multimodal representation learning approaches, creating joint embedding spaces for text, audio, logs, and structured data.
• Enhance speech and audio intelligence pipelines, including speech‑to‑text (STT) and automatic speech recognition (ASR).
• Define evaluation methodologies to measure agent performance in real‑world environments.
• Translate complex interaction signals into structured training objectives for large‑scale models.
• Collaborate with engineering and product teams to move research into production and iterate based on live feedback.

Required profile

• PhD in Computer Science, Machine Learning, AI, Electrical Engineering, or a related field.
• 5+ years of experience in applied AI research or ML systems with production impact.
• Strong scientific rigor combined with practical system‑building skills.
• Excellent written and verbal communication skills in English.

Required skills

• Python programming
• PyTorch
• Hugging Face ecosystem
• Retrieval‑Augmented Generation (RAG) systems
• LLM fine‑tuning
• Reinforcement learning methods (RLHF, DPO, GRPO)
• Representation learning and multimodal embeddings
• Speech‑to‑text (STT) and automatic speech recognition (ASR) pipelines

What we offer

• Fully remote work within a global AI research organization
• Opportunity to shape cutting‑edge agentic and multimodal AI systems
• High‑impact research with direct production deployment
• Collaboration with top‑tier engineering and product teams
• Strong ownership of research projects and rapid iteration cycles