AI engineer in LLM systems

Orbitant; United States; Remote; USD 44,000 - 67,000; Full time

Benefits offered by this job

Remote work (Spain-based)

Training and mentoring

Flexible benefits

Health insurance

Job summary

Orbitant ищет инженеров по искусственному интеллекту (mid и senior) для разработки production-grade систем с LLM. Мы работаем над агентами, RAG и цепочками инструментов, ориентируясь на реальные бизнес-результаты и масштабируемые решения для международных клиентов.

Требуется глубокий опыт Python, владение LangGraph/LangChain и навыками оценки эффективности. Предлагается 100% удалённая работа из Испании, гибкость, обучение и наставничество, а также участие в развитии внутренней AI-платформы

Qualifications

• Опыт разработкиproduction-систем с LLM
• Основной язык Python
• Опыт оркестрации задач через фреймворк (LangGraph, LangChain и др.)
• Опыт проектирования eval-ов и трассируемости LLM
• Понимание prompt-инженерии (структурирование, few-shot, CoT)
• Опыт с Vector DBs (pgvector, Qdrant, Pinecone и др.) и RAG
• Умение сопоставлять затраты и latency между провайдерами/моделями
• Опыт с Claude/Anthropic экосистемой приветствуется
• Знание MCP (Model Context Protocol)
• Крепкий англ. язык для международной команды

Responsibilities

• Проектирование и развёртывание AI-агентов и систем на основе LLM для клиентов
• Вывод проектов от пилота до продакшна: определить и зафиксировать целевые метрики
• Разработка и поддержка RAG-пайплайнов и ingestion-трейсов
• Проектирование архитектур агентных 워флоу (LangGraph) с управлением состоянием
• Интеграция MCP-серверов и связь агентов с ERP/CRM
• Участие в создании внутренней AI-платформы Orbitant
• Проводить ревью кода и наставлять младших коллег
• Проектировать и внедрять практики cost-эффективности и observability

Skills

Python

LangGraph

LangChain

Langfuse

Vector DBs

MCP

Tools

LangGraph

LangChain

Langfuse

Job description

Описание

At Orbitant we are looking for AI Engineers (mid and senior) specialized in building production-grade systems with LLMs, with experience across different types of projects and environments, but above all with a strong desire to keep growing.

Задачи

Design, build, and ship to production AI agents and LLM-based systems for national and international clients — from scoping with the client to deployment and operation. Take projects from pilot to production: define the evaluations that matter, instrument traceability, and ensure that what ships meets the agreed business outcomes. Implement and maintain RAG systems over client documentation and knowledge: chunking strategy, retrieval, reranking, ingestion pipelines. Design multi-step agent and workflow architectures (e.g. with LangGraph) — state management, tool use, error recovery, cost management. Build evals (offline and online) with tools like Langfuse: define metrics, datasets, benchmarks against real outcomes — not against "the demo works". Design and integrate MCP servers and connect agents with existing client systems (ERPs, CRMs, internal tools). Contribute to Orbitant's internal AI-first platform: reusable skills, libraries, templates, and components that accelerate delivery of future projects. Question the "why" before the "how" — collaborate with business profiles to define the outcome before the architecture. Review and improve code (your own and your teammates'), work with legacy code when needed, and maintain high quality standards. Mentor more junior teammates and share learnings (technical sessions, documentation, code reviews). Propose new techniques, tools, and processes when they add real value (not for hype). Be accountable for what you put in production: cost, latency, observability, failures.

Требования

Experience building production systems with LLMs: agents, RAG, generation pipelines, automations. Solid Python as your main working language. Hands-on knowledge of an agent/workflow orchestration framework (e.g. LangGraph, LangChain, etc.). Experience designing and implementing evaluations (evals) and traceability of LLM-based systems (e.g. Langfuse or another observability tool). Real understanding of prompt engineering: structuring, techniques (few-shot, CoT, structured output, tool use). Experience with Vector DBs (pgvector, Qdrant, Pinecone, Weaviate, Chroma...) and RAG patterns (chunking, retrieval, reranking). Ability to reason about trade-offs between providers and models (cost, latency, context window, capabilities) without being locked into a specific API. Anthropic power user: hands-on mastery of Claude Code, Claude SDK, and the Anthropic ecosystem. MCP (Model Context Protocol): designing or consuming MCP servers. Solid engineering practices: clean code, testing, CI/CD, version control, ability to work with legacy code. Level of English that allows communication with international team and clients. "Pilot-to-production" mindset: errors, retries, observability, cost, offline + online evals. Будет плюсом: TypeScript/Node for full-stack integrations and AI-native product backends; Experience with Cloud infrastructure (AWS/GCP/Azure) and knowledge of Docker/Kubernetes; Experience with fine-tuning or lightweight training (LoRA, custom embeddings, reranking); Experience mentoring and helping other people grow; Knowledge of AI security and privacy (PII, prompt injection, jailbreaks, data leakage); OSS contributions, papers, talks, or public AI side-projects

Условия

100% remote work (Spain-based) Top office in Madrid with free attendance Real flexibility Latest generation MacBook Claude license Training and mentoring Flexible benefits with the possibility of health insurance Salary between €40,000 and €60,000 gross per year Referral bonus The chance to build the kind of engineering that gives companies escape velocity, and to reach your own Be part of a spectacular team that can't wait for you to join