Lead AI Engineer & Architect based in Cascais, Portugal.

I design and build reliable AI systems from agentic pipelines and RAG architectures to full-stack GenAI applications at the intersection of technology, governance, and ecological impact.

• Agentic AI & LLMOps — LangGraph orchestration, multi-agent systems, structured outputs, observability (LangFuse, MLFlow)
• RAG pipelines — retrieval optimization, RAGAS evaluation, embeddings, vector databases, knowledge graphs (Neo4j)
• Cloud & DevOps — AWS (CDK, ECS, Lambda, Bedrock, S3, SQS, DynamoDB, API Gateway, CloudWatch), Docker, CI/CD
• Full-stack GenAI — Python backends, Svelte/React frontends, real-time and batch AI pipelines
• Sustainable & ecological AI — decision-making systems for governance, ESG, and regenerative design

• Languages: Python (8+ yrs) TypeScript SQL (6 yrs) Node.js
• GenAI & Agentic Systems (3+ yrs): LangGraph LangChain LlamaIndex Claude OpenAI
• LLMOps & Tooling: RAGAS LangFuse Crawl4AI Pydantic
• Data & Vector Stores (3+ yrs): Neo4j pgvector ChromaDB Supabase
• Cloud (AWS) & DevOps (4-5 yrs): CDK ECS Lambda Bedrock DynamoDB Docker
• Frontend Frameworks (6 yrs): Svelte React

First-author publication — "Exploring a biocentric LLM-based assistant in environmental decision-making with more-than-human representation of the Tagus Estuary"
→ Discovery Sustainability, Springer Nature (2026) 1,716+ accesses

• Capgemini x Volkswagen — Production GenAI commercial electric vehicle sales platform across 10 EU markets, reaching an impacted logistics population of over 150,000. Integrated national lead enrichment registries, multi-language RAG argumentation, and ESG pipelines on 15+ AWS CDK stacks. 95% prompt extraction accuracy; 94% test coverage.
• ITI / LARSyS — Led solo development of multi-language, multi-modal agentic LLM apps for a 6-country industry-academic consortium. 100+ users, 70+ research participants, Springer Nature paper.
• Stealth AI Startup (Técnico Lisboa Spin-off) — Sole AI architect; 2 PoCs to production MVP in 6 months; post-pivot MVP achieved strongest commercial traction to date.