The English-language job board for Spain
91,000+ live jobsAI Agent Engineer
Hybrid in Madrid·Full-time·Added yesterday (last checked 19 Nov)
Our summary
- Work mode
- Hybrid, Madrid
- Contract
- Full-time
- Spanish
- Expected (ad in Spanish)
- Benefits
- Private health insurance, life insurance




Visa options
The ad doesn’t mention sponsorship
Not a refusal, but not a yes. Worth asking in your first message.
Skills
This job advert was originally written in Spanish.
Requirements
What we're looking for
- Advanced Python (typing, async, testing, packaging).
- Real experience (one year or more) building systems with LLMs in production, not just prototypes.
- Practical knowledge of at least two agentic frameworks (LangChain/LangGraph, OpenAI SDK/Agents SDK, PydanticAI, CrewAI...).
- Experience with RAG: vector databases (Qdrant, Weaviate, pgvector, Milvus...), chunking strategies, and retrieval evaluation.
- Experience in the design or implementation of agentic integrations with MCP (servers, clients, or both), understanding tools, resources, and prompts.
- API development with FastAPI and knowledge of Docker.
Nice to have
- Contributions to the MCP ecosystem or experience with the official SDK (mcp in Python or TypeScript).
- Familiarity with MCP security: authorization, prompt injection via tools, tool execution sandboxing.
- Self-hosted deployment and administration of LiteLLM or Langfuse (gateway proxy, key management, virtual keys, budgets, trace export).
- Design of continuous evaluation pipelines: golden datasets, LLM-as-judge, quality regression in CI/CD.
- Experience with at least one AI Gateway (LiteLLM, Portkey, OpenRouter, or similar) or with building custom LLM provider abstraction layers.
- Experience with LLM application observability platforms (Langfuse, LangSmith, Arize Phoenix, or similar), including trace instrumentation in agentic systems.
- LLM inference engineering: vLLM, quantization (GGUF, AWQ, GPTQ), fine-tuning (LoRA/QLoRA), ONNX, self-hosted serving.
- Knowledge of the internal architecture of transformers at a practical level (attention, KV cache, context window management).
- Experience with guardrails and LLM security (prompt injection, output validation, tools like Guardrails AI or NeMo Guardrails).
- Kubernetes and cloud infrastructure for AI service deployment.
- Knowledge of local/open-weights models and when to use them versus closed APIs.
Pay & benefits
What you'll get
Babel, the great way to achieve the success.
Do you want to be part of an expanding, committed, and innovative team that makes history every day? At Babel, we accompany you on your path to success. We believe in the talent of people and we want to empower it by offering you a great working environment based on collaboration and solidarity.
Working at Babel is much more than working at a company; it is joining a team of people with a shared mission and a company model centered on values.
And we do this by promoting diversity, equity, and inclusion, guaranteeing equal opportunities to all candidates, in an environment free of any type of discrimination.
Additionally,
We bet on work-life flexibility. (Point-based vacation system, remote work, and flexibility).
We invest in the training of our Babelievers.
We invest in your physical and emotional health. (Private Medical Insurance/ Life insurance, personal trainer, and wellbeing program).
We also invest in your free time (Nature activities, sports championships, streaming platform subscription (e.g., Netflix))
We participate in the growth of the company and recognize the participation of each person (Benefits Bonus).
We focus on working with the latest technologies and innovative projects, but above all, we accompany digital transformation and bring technology closer to society.
Do you accept the challenge? We are waiting for you!
The role
We are looking for profiles to design, build, and deploy LLM agent-based applications and products in production: multi-agent orchestration, knowledge retrieval (RAG), and high-performance APIs.
What you'll do
- Design and implement agentic architectures (single-agent, multi-agent, workflows vs. autonomous agents) using LangChain, LangGraph, OpenAI Agents SDK, or other frameworks (CrewAI, PydanticAI, AutoGen...).
- Build RAG pipelines: ingestion, chunking, embedding, hybrid retrieval (vector and keyword), reranking, and retrieval quality evaluation.
- Develop APIs and services with FastAPI, including streaming (SSE/WebSockets), concurrency, and conversational state management.
- Integrate tool use and function calling via MCP (Model Context Protocol): design and implementation of custom MCP servers that expose data, APIs, and internal capabilities to agents, and consumption and orchestration of third-party MCP servers.
- Define tool design from the consumer's perspective: granularity, input and output schemas, error handling, idempotency, and security of the exposed tools.
- Implement secure execution environments (sandboxes) for generated code: ephemeral containers, microVMs, network, and resource restrictions.
- Implement agent observability and evaluation with specialized platforms (Langfuse, LangSmith, Arize Phoenix, OpenTelemetry): full tracing of agentic executions, cost and latency analysis per trace, evaluation datasets, automatic scoring, and quality regression detection.
- Manage access to LLM models through AI Gateways (LiteLLM, Portkey, OpenRouter, or similar): routing between providers, fallbacks, rate limiting, budgets per team or application, semantic caching, and centralized cost control.
- Work on inference engineering when the project requires it: serving optimization (vLLM, TGI, Ollama), quantization, continuous batching, KV-cache, model selection, and cost per token.
- Collaborate with product to translate fuzzy requirements into well-defined agentic behaviors.
What is our strategic plan?
We are One Team. We make it happen. We are Unstoppable.
BABEL is a multinational technology consultancy specializing in applying its technological services and knowledge to the digital acceleration processes of its clients, large companies, and public organizations.
We look toward the future! Our Hyperspace 2029 strategic plan promises to be an exciting journey, full of opportunities to grow and develop professionally. Reaching 1 billion in revenue, a challenge that we are sure that with the collaboration and talent of our people, will be another success story that we will write together.
Mission
Develop end-to-end agentic systems, from agent architecture to its exposure as a service, ensuring reliability, scalability, and controlled cost in production.
Attitudes we are looking for
- Architectural pragmatism: you choose the simplest solution that works; you know when a deterministic workflow is better than an agent.
- Obsession with production reliability: agents fail in new ways, and you love anticipating them.
- Continuous learning in a field that reinvents itself every month.
- Autonomy and clear communication with technical and business profiles.
More jobs like this
or browse Madrid·Engineering·Machine learning & AI·Junior·Hybrid·Artificial Intelligence·Jobs of the Future·Community of Madrid·IT Services
