Agentic AI engineer for customer service

SIXT; Lisboa; Hybrid; EUR 70,000 - 110,000; Full time

Benefits offered by this job

2 office days per week

28 vacation days per year

Birthday off

Volunteer day

Private health insurance including fam

Coverflex allowance

Daily meal allowance

Exclusive SIXT rates

Job summary

SIXT ищет специалиста по Data Science в команду трансформации сервиса поддержки клиентов. Ваша работа будет влиять на миллионы пользователей: CARA аутентификация, управление бронированиями и вопросы по политике на более чем 10 рынках и 7 языках.

Вы создадите production multi-agent систему с доступом к платформе бронирования через чат, приложение, WhatsApp и голосовую интеграцию. Вы разрабатываете логику агентов, графы оркестрации, инструменты и оценочные сценарии, обеспечивая безопасность

Qualifications

• Опыт разработки продакшн-решений на Python
• Опыт работы с инструментами LLM (LangGraph/LangChain) и API крупных провайдеров
• Опыт в настройке LLMOps: трассировка, evals, guardrails, CI/CD

Responsibilities

• Разработка и поддержка агентов end-to-end: задач платформа, графы оркестрации и вызовы инструментов
• Определение требований к поведению агентов совместно с бизнес-едром и продакт-менеджерами
• Оптимизация затрат и задержек: маршрутизация между моделями, кэширование промптов
• Мониторинг и аудит в продакшне, сбор метрик и выводs в дашборды
• Сотрудничество с инженерами, дата-сайентистами, MLOps и платформенными командами

Skills

Python

AWS Cloud

LLM tooling

English fluency

Tools

LangGraph

LangChain

OpenAI API

Anthropic Claude

AWS Bedrock

Job description

Описание:

Join our Data Science team and help shape SIXT's Customer Service Transformation. Our AI assistant CARA authenticates customers, manages real reservations, and handles policy questions across 10+ markets and 7 languages. She's a production multi-agent system with write access to SIXT's booking platform, running on chat, app, WhatsApp, and soon voice. Your work will impact millions of customers and deliver measurable business results.

Задачи:

You own agents end to end: prompt, orchestration graph, tools, evals, deployment, dashboard. You write the agent logic and watch your own traces in production You translate business problems into agent capabilities, working with product, business and customer service stakeholders to define tool calls and sub-agents, and you build for real transactions with step-up authentication, guardrails, idempotent case handling, and handover into Salesforce and Genesys You make quality measurable, extending our evaluation harness with LLM-as-judge rubrics, tool-trajectory assertions and prompt A/B suites that gate every release You engineer for cost and latency as much as accuracy: multi-provider model routing with fallbacks, prompt caching, small models where small models win You collaborate across the stack with software engineers, data scientists, MLOps and platform teams, balancing bespoke agent logic against reusable platform capability You raise the floor around you by sharing agentic and LLM engineering knowledge with colleagues and helping other teams integrate Agentic AI into their products

Требования:

You're an engineer first with strong, hands-on software engineering, real production ownership, excellent Python skills, and solid AWS cloud experience You have built tool-calling agents with LangGraph/LangChain or equivalent frameworks and shipped them to actual users rather than keeping them in notebooks You're hands-on with frontier and open models like Anthropic Claude, OpenAI, AWS Bedrock, and open-source models such as Llama, with a clear point of view on when to use which You treat prompt engineering as an engineering discipline: versioned, reviewed, evaluated, A/B tested You know and have applied LLMOps: tracing, evals, guardrails, token and latitude budgets, continuous integration and deployment You're extremely motivated, passionate about learning, bring a genuine can-do attitude, communicate fluently in English, and appreciate cultural diversity Будет плюсом: MCP, real-time voice AI (WebRTC, SIP, streaming STT/TTS), event-streaming pipelines, TypeScript/React, or contact-centre platforms like Genesys and Salesforce add value

Условия:

• Two office days a week, flexible hours, and modern spaces built for both deep focus and team collaboration – work the way that brings out your best
• Enjoy 28 vacation days per year, an additional day off on your birthday, one volunteer day annually, and private health insurance that also covers your spouse and children free of charge
• Get a yearly Coverflex allowance and a daily meal allowance to make everyday life a little easier
• Get exclusive rates on all SIXT services, with special discounts extended to your family and friends.
• Because great perks are even better when you can share them