EU remote
AI & Data Engineer (all genders)
About this role
ABOUT THE ROLE You'll be the senior engineering anchor of our Lisbon AI hub. You'll ship production-grade agentic AI systems for enterprise clients (orchestration, RAG, evals, deployment, observability) AND own the data plumbing they depend on (dbt models on a warehouse layer over HubSpot and client data). The AI Team today has more qualified AI requests in the pipeline than engineers capable of taking them from prototype to production.
You change that. WHAT YOU'LL DO - You'll ship 3+ production agentic AI systems to enterprise clients within 12 months (deployed, signed off, in active use) - You'll build and maintain our reusable agent framework: tool-calling layer, prompt management, evals harness, observability, usable by future AI Engineers across all sites - You'll build and operate the internal data warehouse layer (dbt models, source freshness checks, tested transformations) for both client AI engagements and internal analytics - You'll establish our evals and data quality bar: every shipped agent gets documented evals, drift monitoring, and a rollback plan; every shipped pipeline gets tests and freshness SLAs - You'll maintain ≥ 85% billable utilisation across active engagements - You'll co-author 2+ public-facing technical assets (blog post, webinar, conference talk) on our AI or data engineering approach - You'll contribute to AI proposal and scoping conversations as the technical authority alongside Sales and Advisory WHAT YOU BRING Must-haves - Direct experience building and shipping agent systems using tool-calling LLMs (Claude, GPT, Gemini) to production.
You understand tool schemas, multi-step reasoning, error handling, retries, observability - Hands-on data engineering experience with dbt (or comparable), modeling CRM or B2B data into a clean dimensional layer, operating a warehouse (BigQuery, Snowflake, Postgres, DuckDB, or similar) under change control - Production RAG experience with vector stores (Pinecone, Weaviate, pgvector), hybrid search, re-ranking, and chunking strategies that actually work for enterprise documents - Evals and data-test discipline.
You don't ship without them. Automated eval suites (golden datasets, LLM-as-judge, regression tests) and dbt-style data tests treated as production code, not notebooks - Production engineering hygiene: maintainable Python or TypeScript, tests, types, CI, deployment, cloud fluency (AWS, GCP, or Azure), Docker, basic IaC - Cost and latency awareness. You read token cost and warehouse cost like an SRE reads CPU. You know when to cache, when to fine-tune, when to switch models or pre-aggregate data - Native-level Portuguese plus clear business English.