About PermutPermut is the leading conversational AI platform for brands to build, optimize and scale AI agents across every customer channel. +100 amazing brands chose Permut in our first year. Our ambition is to become the platform that powers how brands serve, engage and sell through AI.

About the role
We’re looking for a Research Engineer to help build the intelligence behind Permut’s next generation of AI agents. You’ll work on how agents understand intent, use brand knowledge, select tools and execute actions. Your work will combine model adaptation, data curation and rigorous evaluation, with a direct path from experiments to production. Our ambition is to develop proprietary models and learning systems grounded in customer conversations and their outcomes. You’ll help establish the research foundations: which data matters, how we measure progress and how agents improve with experience. What you’ll do
• Adapt, fine-tune and improve models for conversational tasks including product discovery, customer support and tool use.
• Build high-quality datasets from conversations, brand knowledge, agent actions and human feedback.
• Design benchmarks that measure factual accuracy, task completion, tool execution and conversation quality.
• Investigate failures across prompting, retrieval, context management and model behaviour, then design experiments to address them.
• Explore post-training approaches and feedback loops that improve agent performance over time.
• Work closely with engineering to deploy successful experiments and measure their impact on quality, latency, cost and reliability. What you bringWe care about your ability to solve difficult problems and produce convincing evidence. Show us the models, experiments, projects or open-source contributions you’re proud of,



and explain what you learned. Ideally, you bring:
• 3+ years of experience in AI or machine learning, including implementing, training or improving models.
• Strong Python skills and experience with frameworks such as PyTorch and the Hugging Face ecosystem.
• Hands-on LLM experience, including fine-tuning, dataset preparation and evaluation.
• Strong experimental judgment: You can establish useful baselines, isolate variables and assess whether improvements hold up in real use.
• Experience improving data quality: You understand sampling, labeling, coverage and data leakage.
• Production awareness: You consider latency, cost, reliability and monitoring alongside model quality.
• Autonomy and clear communication: You can turn an open question into a research plan and work with engineers to ship the result. Experience with conversational AI, retrieval, tool use, preference optimization or reinforcement learning is a strong plus. Publications are welcome, alongside evidence of strong independent or commercial work. How we work
• Move fast: Short experimentation cycles, direct feedback and clear priorities.
• Own the outcome: Help define the problem, run the experiments and bring successful work into production.
• Stay close to customers: Real conversations and customer needs shape our research priorities.
• Be rigorous: Reproducible experiments, meaningful evaluations and reliable production behaviour guide our decisions. What you get
• The opportunity to shape Permut’s research direction and proprietary intelligence.
• Real customer problems and a direct path from research to production.
• Close collaboration with the founders and engineering team.
• Room to grow across model development, data, evaluation and applied AI.
• The tools, resources and autonomy to do your best work.

📌 Research Engineer (España)
🏢 Permut
📍 España