Summary: This role involves designing and optimizing prompts for an AI agent that automates debt collection, focusing on improving its interactions and problem-solving abilities. Highlights: 1. Shape AI intelligence through prompt design for real-world automation. 2. Work with LLMs beyond basic prompting in a fast-moving AI startup. 3. Directly impact how businesses communicate and operate through AI. **About Kleva** Kleva (kleva.co) is an AI agent that automates debt collection. We’re building intelligent systems that handle real customer conversations — not just chatbots. **About the role** Your job is to make Kleva think better. You’ll design and optimize the prompts that shape how Kleva interacts with customers, solves complex cases, and adapts to different scenarios. **Responsibilities** * Design, test, and refine prompt architectures for diverse customer use cases * Analyze model outputs to improve accuracy, tone, and outcome success * Collaborate with engineers to productionize prompt workflows * Monitor and document prompt performance metrics **What we actually look for** * Strong understanding of collections, lending, or financial services (BPO experience, fintech, credit ops, or call center management). * Comfort with LLMs beyond basic prompting: system prompts, few\-shot learning, tool use, constraints, output validation. * Spanish fluency. Most of our calls are in Spanish. If you can't prompt in Spanish, you can't do this job. * Ability to think in structured logic. Collection calls are decision trees with emotional variance, not freeform conversation. * High tolerance for iteration and failure. Most prompts break on live calls before they work. **Why join Kleva** Join a fast\-moving AI startup tackling real\-world automation challenges. Your work will directly impact how businesses communicate and operate through AI. If you’re obsessed with how language shapes intelligence and want to build smarter AI agents, we’d love to hear from you. **Requirements** * Proven experience working with large language models (OpenAI, Anthropic, etc.) * Deep understanding of prompt design, context management, and evaluation * Strong analytical and writing skills * Comfortable experimenting and iterating quickly