One project brain
Every drawing, RFI, permit, email, and decision — searchable from a browser. Source-traceable. Your team’s data, not OpenAI’s training set.
Project memory · Portugal · EN / PT
Your Portugal project is leaking time, files, and decisions.
I install AI-augmented project memory for international developers, architects, and studios running projects in Portugal — so your team stops losing six weeks to “wait, did we send that to the engineer?”
Book a free 15-min call See what’s builtCivil engineer + architectural designer + developer · Based in Portugal · English & Portuguese
01 — The problem
If any of this sounds like your week, keep reading
- RFIs are scattered across WhatsApp, email, and three Dropbox folders
- Every time a project manager leaves, you lose half the context
- The Câmara is asking for a document your architect insists was already sent — but nobody can find it
- You’ve paid four consultants and still can’t get one coherent picture of where the project stands
- You want AI to help, but you don’t trust general ChatGPT with your client files
Most AEC firms today run on Dropbox + WhatsApp + email. The information exists. The retrieval doesn’t.
02 — What you get
One project brain for your AEC projects
Not a chatbot. A project-memory system with deterministic curation, scheduled briefings, multi-agent retrieval, and auditable governance.
2.1
2.2
Plain-English queries
“What did the structural engineer say about the foundation in March?” — answered in seconds, with sourced citations to the original PDF.
2.3
Onboarding in an afternoon
A new coordinator or consultant gets context that used to take a month. The audit trail satisfies licensing, insurance, and disputes.
03 — Real proof
What’s actually built
Systems in production, delivered, or in active development — each documented as a full case study.
In production
Project Memory System
A multi-agent context system I run daily on 305 MB of my own data. 7,571 lines of Python, 12 specialized AI skills, 11 vaults, scheduled connectors. The deployed prototype behind every Project Brain Build.
Case study →Ran end-to-end
Green Score — Embodied Carbon
2-agent CrewAI + Claude system: IFC model + spec book + takeoff in, defensible carbon report out. Real 186 m² residence: 102,410 kg CO₂e, 5 mitigation actions, −17.7% at 2.2% of project cost.
Case study →Delivered 2026
Bret & Rachel — BIM Rebuild
Private commission: a 2D CAD drawing set rebuilt as a full parametric Revit model and rendered photoreal — so the clients could see the house before it was built.
Case study →Team thesis · Aug 2026
BIM Acoustic Coordination
7-agent system verifying acoustic compliance (RRAE / DL 96/2008) on IFC models — Zigurat master’s thesis, team of five, real Portuguese test project.
Case study →Solo R&D
ABNT Compliance Audit
5-agent CrewAI crew auditing memorial, budget, and IFC against three ABNT standards — then rewriting each document with corrections traced to norm clauses.
Case study →Research
EU AI Act Governance
A deployable 7-directive governance protocol for multi-agent AEC systems under Regulation 2024/1689 — full application 2 August 2026.
Case study →04 — Worked example
How an engagement plays out
Illustrative engagement — fictional clients and plot. Real workflow, real regulations, real deliverables.A UK couple was at soft-offer stage on a €490,000 rural plot in the Lagoa municipality, with a UK-architect-drawn villa already sketched. The plot looked clean. Nobody could answer the real question — is what they want to build permittable on what they’re about to buy?
A Pedido de Informação Prévia filed with the Câmara Municipal de Lagoa came back with three constraints the original team hadn’t accounted for: a buildable footprint capped at 192 m² (not the 240 m² in the design), a 50-metre fire-defence buffer, and a mandatory ETAR adding ~€18,000 to the build.
- The binding PIP response, archived as the project’s foundational document
- A re-negotiated purchase at €452,000 — €38,000 saved before the deed
- A reshaped brief, UK architect and Portuguese arquiteto on one bilingual constraint sheet
- An issued alvará de obras 22 weeks from intake — against a 9–14 month Algarve baseline
- The project memory itself, handed over as the operational record into construction
€60,000 in avoidable cost caught at PIP stage. Six months of delay avoided.
Read the full worked example →
05 — How we work together
Three ways to start
Audit
€500 founding rate — €750 thereafter
90-minute call, written diagnostic of where your project memory is leaking, 30-day fix plan. Delivered in five business days.
Details →Project Brain Build
€4,500 founding rate — €7,000 thereafter
In 3–4 weeks I install a working, AI-queryable project memory for one project or studio — ingest, index, query layer, team training, 30 days support.
Details →Curator Retainer
from €2,000/month
Your project brain stays alive without your team maintaining it. Weekly ingest, prompt tuning, monthly review. Capped at four clients. Offered after a Build.
Details →Why the rates are this low: I’m taking the first two paying clients at founding rates in exchange for case-study rights. Read the honest version →
One 15-minute call
No pitch deck. No script. I diagnose your specific situation on the call. If it’s not a fit, I’ll tell you on the call.
Book the call