01 / Agents
Agents & MCP servers
Multi-step agents with tool calling, structured outputs and human-in-the-loop approvals. MCP servers that plug them into real business tools and data.
LangChainLangGraphMCPTool calling
I BUILDAGENTSTHAT SHIP
Previously at·Rolex·Helvetic Payroll·Grow Surely·42 Lausanne·Business School Lausanne·AI builder·Switzerland·Previously at·Rolex·Helvetic Payroll·Grow Surely·42 Lausanne·Business School Lausanne·AI builder·Switzerland·
Live demo
Ask my agent.
Recruiters ask me the same ten questions, so I built an agent that knows the answers. It runs on the same architecture as my own assistant. When a question needs me, it can message me or put a call on my calendar.
Has Lucca put agents in production?
Yes. At Helvetic Payroll, for Rolex, he shipped GenAI assistants and MCP servers wired into internal tools. Want me to book you a call with him?
lucca-agent · onlinev1
What did you build at Rolex?Have you shipped MCP servers?Are you open to new roles?Book 30 min with Lucca
search_profile()Answers from my experience, projects and stacksend_lucca_a_message()Drops your note straight into my inboxbook_meeting()Finds a free slot on my calendar and books it
// "just make it work in prod." \\
What I build
for teams like yours.
02 / Retrieval
RAG & document pipelines
Retrieval systems and vision-language pipelines that turn thousands of messy enterprise files into validated, structured data.
RAGpgvectorAzure AI SearchVLMs
03 / Automation
Workflows that run themselves
Automations that connect research, enrichment, CRM and outreach, mixing LLM output with plain business rules where each one fits.
n8nWebhooksHubSpotAPIs
Built in public · my own agent harness runs this site's agent
Stack
PythonTypeScriptLangChainLangGraphMCPFastAPINext.jsPostgrespgvectorDockerGCP / Vertex AIAzure AI SearchGeminin8nVercel AI SDKGitLab CI/CD
The record
Three roles.
One habit: shipping.
AI Engineer ConsultantHelvetic Payroll (for Rolex) · Switzerland · Sep 2025 – Jun 2026+
- Built production GenAI assistants and multi-step agent workflows in Python and JavaScript, owning implementation, debugging and iteration end to end.
- Built MCP servers and REST API integrations connecting AI agents to business tools and data, using tool calling, structured outputs and reusable components.
- Built RAG systems and AI-powered internal tools, turning ambiguous business needs into working automations with stakeholders.
Digital Innovation InternRolex · Geneva · Mar 2025 – Aug 2025+
- Built a Python document-processing pipeline handling ~6,000 enterprise files, automating extraction, transformation and validation with vision-language models and structured outputs.
- Benchmarked AI models and providers on quality, reliability, trade-offs and business fit to guide tooling decisions.
- Evaluated emerging AI tools and turned findings into implementation recommendations for business teams.
Sales EngineerGrow Surely · Remote · Feb 2024 – Mar 2025+
- Built and launched 8+ automation workflows with n8n, HubSpot, APIs and enrichment tools, covering research, enrichment, outreach and follow-up.
- Built AI-assisted workflows combining LLM outputs, structured rules and business logic to automate repetitive operations and personalization.
- Designed scoring and prioritization logic and iterated on it directly with clients.
42 LausanneComputer Science & Software Development2024 – 2026
Business School LausanneBachelor, Business Administration2021 – 2023
By the numbers
Small numbers, real systems. Everything here ran on actual company data.
~6,000
Files through one pipeline
Enterprise documents extracted, transformed and validated with vision-language models at Rolex.
8+
Workflows launched
Production automations across research, enrichment, CRM and outreach at Grow Surely.
MCP
Servers in production
Agents connected to real business tools through MCP servers and REST integrations.
4
Languages
English, French, Portuguese and Spanish, so the stakeholder call doesn't need a translator.
Biz + Eng
Two backgrounds
A business degree and 42 Lausanne. I can read the P&L and the stack trace.
0 → prod
End-to-end ownership
Scoping, building, debugging and iterating, not just the demo.
General
What do you actually build?+
Agents that do real work: assistants with tools and approvals, MCP servers that connect them to business systems, RAG over company data, and the automations around all of it. Most recently for Rolex, through Helvetic Payroll.
What's your stack?+
Python and TypeScript. LangChain/LangGraph for agents, FastAPI and Next.js around them, Postgres + pgvector for memory and retrieval, Docker, and GCP or Azure depending on the client. n8n when a workflow tool is the right answer.
Where are you based?+
Switzerland. I work in English, French, Portuguese and Spanish.
Can I just ask your agent?+
Yes, that's the point. It knows my experience and projects, can pass a message to me, and can book a call on my calendar.
How do I reach a human?+
Book a slot with the button on this page, or email luccafmourao@gmail.com. I answer quickly.