Shine is the financial copilot for entrepreneurs and small business owners. Founded by serial entrepreneurs Rico Andersen and Martin Hegelund, Shine is a leading European fintech unicorn on a mission to restore the joy of running a business, by ending wasted time on financial admin. Shine offers a connected solution for invoicing, accounting, payroll, business accounts, payments, and financing, meaning business owners can focus their energy on growing a healthy business, not held back by manual admin. Part of something bigger Today we're part of Cegid, a European leader in cloud software for finance and accounting. Together we're building Europe's leading financial copilot for small businesses and their accountants. Shine already supports more than 400,000 small businesses. As part of Cegid, we now reach over one million small businesses and 15,000 accountants across Europe. We're a multicultural team working from France, Germany, Denmark and the Netherlands, Spain,Portugal. Your hiring experience matters Just as we respect our customers' time, we respect yours. Your experience with Shine and Cegid should feel simple, transparent and genuinely supportive. If this sounds like somewhere you want to grow, we'd love to hear from you. The Advisor Production team at Shine The Advisor Production team is building AI systems that automate accounting production for accounting professionals. Today, our agentic pipeline is already live in production: it analyzes invoices and proposes double-entry accounting entries for CPA validation. Every proposal is reviewed by a human accountant, making correctness, reliability, and traceability more important than raw speed. We're now expanding the platform to additional countries, starting with Denmark. Our engineering culture values strong conventions, high ownership, and production-grade AI engineering. Our codebase is a TypeScript/NestJS Nx monorepo with rigorous error handling, structured LLM outputs, and an evaluation framework that ensures every change is measurable. Your role as an AI Backend Engineer Automate accounting production for Advisors by turning source documents into proper bookkeeping. As an AI Backend Engineer, you'll own and improve production AI agents that solve real accounting problems. You'll work across the full stack of agent development—from prompt engineering and evaluation to backend services, orchestration, and production observability. Your responsibilities will include: Building and maintaining AI agents for document parsing, VAT resolution, account selection, and other accounting workflows using structured outputs and clear contracts. Designing and operating production-grade AI workflows with robust orchestration, tool calling, retries, and failure handling. Integrating LLM APIs with typed schemas while optimizing for quality, latency, and token cost. Writing integration tests and evaluation suites to continuously measure and improve agent performance. Instrumenting the production pipeline with traces, structured logs, and monitoring to quickly diagnose issues. Extending our multi-country accounting logic without duplicating business rules. Collaborating closely with Product Managers and CPAs to understand accounting edge cases and continuously improve agent quality. Contributing to the CPA feedback loop by transforming manual corrections into reusable rules and smarter automation. Helping shape the architecture of our AI platform as we expand across Europe. 📍 Job located in Berlin, Copenhagen or Amsterdam with the possibility of working remotely up to two days per week. About you You have 4+ years of experience building and operating production backend systems. You're a strong backend engineer with TypeScript / Node.js (NestJS is a strong plus). You've built or operated AI agent workflows in production —including orchestration, tool calling, structured outputs, retries, and error handling—not just LLM prototypes. You have hands-on experience integrating LLM APIs (OpenAI, Anthropic, or similar) and think carefully about schemas, latency, costs, and failure modes. You care deeply about software quality through testing, observability, and evaluation frameworks. You write rigorous, strongly typed code with disciplined error handling. You're comfortable moving between prompt engineering, backend services, queues, databases, and infrastructure without relying on handoffs. You have strong ownership, excellent written communication, and enjoy collaborating with engineers, product managers, and accounting experts. You're curious but skeptical about LLMs—you see them as probabilistic systems that must be constrained, measured, and continuously improved. You're detail-oriented and understand that in accounting, correctness matters. Nice to have Experience with event-driven architectures (SQS or equivalent). PostgreSQL and pgvector. Nx monorepos. Datadog, Langfuse, or similar observability platforms. Fintech or accounting domain experience. Experience designing LLM evaluation methodologies and prompt regression testing. Language requirements Professionally fluent in English. Our stack TypeScript, NestJS, Nx monorepo, PostgreSQL + pgvector, TypeORM, SQS, OpenAI APIs, Langfuse, Datadog, neverthrow, structured LLM outputs, integration testing, and an evaluation harness. Equal Opportunity Employer We follow the principle of equal treatment to consider all job applicants and do not discriminate based on their gender, sexual orientation, color, racial or ethnic origin, religion, disability, etc. as per applicable law. Our recruitment process 1️⃣ Recruiter interview: A first conversation to get to know each other and answer your questions. 2️⃣ Hiring Manager interview: A deeper discussion about your experience, technical approach, and the role. 3️⃣ Technical case study: A practical exercise followed by a discussion of your solution and thought process. 4️⃣ Personality assessment: A short assessment to help us understand how you work and collaborate.