Any questions?
2IIA
2IIA Assistant
Hello!
I'm the 2IIA assistant.
What are you looking for?
8 SHORT COURSES · 22 DAYS

Architecture & AI

AI courses for architects and urban planners

Eight short courses to design differently with artificial intelligence: generative AI, territorial data, BIM and digital twins, generative design, AI agents, regulation and reuse. For architects, urban planners, landscape architects, engineers and clients.

DESIGNED WITH THE ARCHITECTURE & INTELLIGENCES CHAIR A 2IIA partner
8
short courses
1 to 5 days
7 to 35 hours
100%
online, live
12
participants max. (most courses)
Overview

8 courses to design differently

Each course belongs to one theme: augmented design, resources and ecosystems, AI for architecture, or a cross-cutting topic. Take a single course or a full pathway.

No.CourseDurationThemeLevel
F1Generative AI for architectural design2 days (14 h)Augmented designBeginner
F2AI- and GIS-augmented territorial diagnosis3 days (21 h)Resources & ecosystemsBeginner
F3Digital twins and BIM for the sustainable city3 days (21 h)Augmented designIntermediate
F4Python and data science for architects and urban planners5 days (35 h)AI for architectureBeginner
F5Computational design and generative algorithms3 days (21 h)Augmented designIntermediate
F6Data, ethics and the AI Act for the city1 day (7 h)Cross-cuttingAll levels
F7AI agents and automation in the architecture firm2 days (14 h)AI for architectureIntermediate
F8AI for reuse and the circular economy in construction3 days (21 h)Resources & ecosystemsIntermediate
Total8 courses22 days (154 h)
Recommended pathways

Where to start?

Three sequences depending on your goals. F1 is the ideal entry point to discover AI in an architecture firm.

A firm starting with AI

Bring AI into design and firm organisation, while staying compliant.

  1. F1 · Generative AI
  2. F7 · AI agents in the firm
  3. F6 · Ethics and the AI Act

Territory and data

Use data for diagnosis, planning and urban management.

  1. F4 · Python and data science
  2. F2 · Territorial diagnosis
  3. F3 · Digital twins and BIM

Low-carbon design

Optimise form and design with reclaimed resources.

  1. F5 · Generative design
  2. F8 · Reuse and circular economy
Detailed programmes

The 8 courses, day by day

All courses are taken online, in a virtual classroom with the trainer, with hands-on workshops on real architecture and urban planning cases.

F1 · Augmented design

Generative AI for architectural design

Use generative AI tools at every project phase, write structured prompts, assess outputs critically and manage the legal risks. No coding required.

Duration 2 days (14 h)Level BeginnerGroup 12 max.Language French or English
LLMsImage generatorsControlNet3D and rendering tools

Objectives

  • Choose the right generative AI tool for each project phase: text, image, 3D, video
  • Write structured prompts for the brief, sketch, mood and design statement
  • Assess AI outputs critically: bias, formal clichés, structural coherence
  • Identify legal risks: copyright, client data confidentiality, mandatory notices

Deliverables

  • A reusable prompt notebook for the firm
  • A board of variants for your own project

Audience and prerequisites

Architects, landscape architects, designers and project managers in firms. No coding skills; bring a current project if possible.

SessionContent
Day 1 am · 3.5 hOverview of generative AI (LLMs, images, 3D and rendering); a simplified explanation of diffusion models
Day 1 pm · 3.5 hApplied prompt engineering: brief, site analysis, references and design statement, in guided exercises
Day 2 am · 3.5 hFrom sketch to render: sketch-to-image (ControlNet), mood and material variants, consistency across views
Day 2 pm · 3.5 hWorkshop on a real project with group critique; legal and ethical framework, a usage charter for the firm

Methods: 30% theory, 70% practice on real project examples. Assessment: graded practical exercise at the end of day 2 and a 15-question quiz.

F2 · Resources & ecosystems

AI- and GIS-augmented territorial diagnosis

Build a territorial diagnosis from open data and QGIS, create vulnerability indicators and read land cover through AI classification of satellite images.

Duration 3 days (21 h)Level BeginnerGroup 15 max.
QGISOpen dataSentinel-2AI classification

Objectives

  • Structure a territorial diagnosis with a systemic approach: ecological, social, economic and spatial components
  • Find and combine relevant open data in QGIS
  • Build and map vulnerability indicators
  • Read land cover and urban dynamics through AI-assisted classification
  • Present a clear diagnosis with scenarios and recommendations

Deliverables

  • An atlas of 5 to 8 maps
  • A systemic strengths / weaknesses diagram

Audience and prerequisites

Urban planners, architects, local authority staff, environmental consultancies, developers. Basic spreadsheet skills; no GIS knowledge needed.

SessionContent
Day 1 am · 3.5 hA systemic approach to territories: interactions, feedback loops, symptoms
Day 1 pm · 3.5 hOpen data sources (IGN, INSEE, Géorisques, Corine Land Cover, DVF): quality and limits
Day 2 am · 3.5 hGetting started with QGIS: layers, joins, spatial queries, proximity analysis
Day 2 pm · 3.5 hVulnerability indicators (heat islands, flooding, access to amenities, deprivation), weighting, synthesis maps
Day 3 am · 3.5 hAI and satellite imagery: land-cover classification, detecting soil sealing, with Sentinel-2 and QGIS plugins
Day 3 pm · 3.5 hGroup mini-diagnosis on a test territory, then a 15-minute presentation per group

Methods: guided work on prepared datasets, with a running case on a town in the Paris region. Assessment: 50% practical work, 50% mini-diagnosis presentation.

F3 · Augmented design

Digital twins and BIM for the sustainable city

Understand BIM, CIM and digital-twin maturity levels, apply ISO 19650 and open formats, then assemble a district-scale digital twin linked to live data to inform decisions.

Duration 3 days (21 h)Level IntermediateGroup 12 max.
BIMIFCCityGMLCityJSONIoT

Objectives

  • Distinguish BIM model, CIM and digital twin, and what each brings to decision-making
  • Apply ISO 19650 and open exchange formats (IFC, CityGML, CityJSON)
  • Assemble a simple district digital twin linked to dynamic data (energy, sensors)
  • Use simulations (energy, comfort, usage) to inform a collective decision

Deliverables

  • A district digital twin
  • A scenario comparison note (2 pages)

Audience and prerequisites

BIM managers, architects, public and private clients, asset managers. Basic BIM concepts (model, IFC); experience with Revit or Archicad recommended.

SessionContent
Day 1 am · 3.5 hFrom data to information to decision; BIM, CIM and digital-twin maturity levels; feedback from cities and asset managers
Day 1 pm · 3.5 hISO 19650, BIM agreements, IFC / CityGML interoperability, model quality control
Day 2 am · 3.5 hBuilding a district twin: importing the model and urban context, georeferencing
Day 2 pm · 3.5 hDynamic data: consumption, IoT sensors (temperature, CO2, occupancy), dashboards
Day 3 am · 3.5 hSimulations: sunlight, summer comfort, energy performance, user flows
Day 3 pm · 3.5 hDecision workshop: comparing scenarios, then a presentation to a simulated steering committee

Methods: demonstrations, then guided practice on a supplied model or your organisation's own. Assessment: continuous assessment of exercises and a final presentation.

F4 · AI for architecture

Python and data science for architects and urban planners

Learn Python in Jupyter, process tabular and geographic data with Pandas and GeoPandas, produce maps, train a first model with scikit-learn, and code with an AI assistant while checking what it produces.

Duration 5 days (35 h)Level BeginnerGroup 12 max.
PythonJupyterPandasGeoPandasFoliumscikit-learn

Objectives

  • Write simple Python scripts in Jupyter to automate repetitive tasks
  • Clean, combine and analyse tabular and geographic data (Pandas, GeoPandas)
  • Produce interactive maps and charts for study outputs
  • Train and interpret a first machine learning model applied to the city
  • Use an AI assistant to write and fix code, and check the result

Deliverables

  • A documented project notebook
  • A library of reusable functions

Audience and prerequisites

Architects, urban planners, landscape architects and study officers who want to process their own data. No programming experience; spreadsheet logic is enough.

SessionContent
Day 1 · 7 hPython basics (variables, lists, loops, functions); the Jupyter environment, good practice, an AI coding assistant
Day 2 · 7 hPandas: importing CSV and Excel files (DVF, INSEE, energy consumption), cleaning, aggregation, joins
Day 3 · 7 hGeoPandas and mapping: shapefiles, GeoJSON, projections, spatial joins; interactive maps with Folium and Kepler.gl
Day 4 · 7 hMachine learning with scikit-learn: regression (price, consumption), clustering (urban fabric typologies); validation and bias
Day 5 · 7 hPersonal project on a dataset of your choice, supported by the trainer; 10-minute presentation

Methods: "one concept, one exercise"; all notebooks are provided and remain available after the course. Assessment: daily self-correcting exercises and the day 5 project presentation.

F5 · Augmented design

Computational design and generative algorithms

Translate an architectural intention into rules and parameters, model in Grasshopper, optimise against several criteria and couple the generative model with AI to sort thousands of variants quickly.

Duration 3 days (21 h)Level IntermediateGroup 12 max.
RhinoGrasshopperLadybugWallaceiGalapagos

Objectives

  • Translate an architectural intention into rules and parameters
  • Build a parametric model in Grasshopper, from geometry to contextual data
  • Run a multi-criteria optimisation (solar, views, material, areas) and read a Pareto front
  • Couple the generative model with AI to sort and assess thousands of variants quickly

Deliverables

  • A documented parametric script
  • A board of optimised variants

Audience and prerequisites

Architects, designers, structural and façade engineers. Experience with a 3D modeller (Rhino preferred); no programming experience needed.

SessionContent
Day 1 am · 3.5 hGeometry, rules and the history of science, from stereotomy to generative algorithms: the algorithm as a design material
Day 1 pm · 3.5 hGrasshopper: flow logic, lists and data trees, parameters; a first generative façade model
Day 2 am · 3.5 hContextual data: Ladybug (sunlight, radiation), views, regulatory envelopes
Day 2 pm · 3.5 hMulti-criteria optimisation (Wallacei, Galapagos): defining objectives, exploring solutions, reading trade-offs
Day 3 am · 3.5 hCoupling with AI: surrogate models that predict performance without full simulation; assisted variant sorting
Day 3 pm · 3.5 hShort project: optimised pavilion or façade; export to digital fabrication (cutting, 3D printing); critical review

Methods: short theory inputs, then intensive practice; all scripts are provided and commented. Assessment: quality and readability of the script, relevance of the optimisation criteria.

F6 · Cross-cutting

Data, ethics and the AI Act for the city

Assess the risk level of an urban AI system, connect the AI Act and GDPR, lay the foundations of data governance, and practise on real local authority cases.

Duration 1 day (7 h)Level All levelsGroup 20 max.
AI ActGDPRDPIAOpen data

Objectives

  • Assess the risk level of an urban AI system under the AI Act
  • Know the obligations already in force and the timeline for the rest
  • Connect the AI Act and GDPR for projects using sensors, cameras or personal data
  • Lay the foundations of data governance: openness, sovereignty, sobriety

Deliverables

  • An AI Act and GDPR self-assessment grid for an urban project
  • A template AI usage charter

Audience and prerequisites

Local elected officials and directors, firm leaders, developers, clients, IT directors. No prerequisites.

SessionContent
Morning · 1.5 hThe AI Act's risk-based approach: prohibited practices (social scoring, certain biometric identification in public spaces), transparency obligations, timeline
Morning · 2 hGDPR and AI in urban space: algorithmic video surveillance, sensors, digital twins, impact assessment (DPIA), the role of the CNIL
Afternoon · 1.5 hData governance: open data, sharing platforms, public-private partnerships, hosting sovereignty, digital sobriety
Afternoon · 2 hCase workshops: smart video surveillance, parking management, a municipal digital twin, scoring of aid applications; each group assesses the risk and proposes safeguards

Latest update. The EU "digital omnibus" regulation on AI, in force since July 2026, postpones the high-risk obligations (to 2 December 2027 for Annex III and 2 August 2028 for Annex I). Obligations already in force remain, including prohibited practices and AI training for staff. The regulatory content is updated before each session.

Methods: short talks, real case studies, a qualification grid given to each participant. Assessment: a 20-question quiz and a workshop presentation.

F7 · AI for architecture

AI agents and automation in the architecture firm

Map the firm's processes, build a document assistant (RAG) on your own documents, set up no-code agents with human validation, secure client data and leave with an adoption roadmap.

Duration 2 days (14 h)Level IntermediateGroup 12 max.
RAGn8nMakeConfigurable assistants

Objectives

  • Identify the firm's automatable tasks and estimate the time saved
  • Build a document assistant (RAG) on the firm's documents: technical standards, planning rules, standard specifications, project archives
  • Set up a simple agent that chains several steps, with human validation
  • Put confidentiality and security rules in place for client data
  • Write an adoption roadmap for the firm

Deliverables

  • A working internal assistant prototype
  • A six-month adoption roadmap

Audience and prerequisites

Partners and directors of firms, project managers, administrative managers, construction cost consultants. Everyday use of a conversational assistant; F1 recommended but not required.

SessionContent
Day 1 am · 3.5 hMapping the firm's processes (regulatory watch, competition entries, site reports, specifications, notices); prioritising by gain and risk
Day 1 pm · 3.5 hRAG principles (chunking, indexing, retrieval, source citation); building an assistant on a regulatory corpus and archives; reliability tests
Day 2 am · 3.5 hAgents and automation: no-code tools (n8n, Make) and configurable assistants; example: competition file → analysis table → draft technical statement
Day 2 pm · 3.5 hSecurity, confidentiality, sovereign hosting, GDPR compliance; change management; individual roadmap presented to the group

Methods: each participant works on a real case from their firm, with anonymised documents where needed. Assessment: prototype demonstration and roadmap review.

F8 · Resources & ecosystems

AI for reuse and the circular economy in construction

Search, characterise and design from scattered, limited stock. Reuse reverses the usual logic: you design with heterogeneous, dispersed resources in small quantities. AI helps at three moments: finding the stock, defining the resource, designing with it.

Duration 3 days (21 h)Level IntermediateGroup 12 max.
PEMD diagnosisINIES / FDESComputer visionGrasshopper

Objectives

  • Search for and aggregate stock from heterogeneous sources: PEMD diagnoses, reuse platforms, client inventories, site surveys
  • Structure a consistent resource database from disparate data (photos, PDFs, spreadsheets, 3D scans)
  • Characterise elements with AI (vision, text extraction, estimation): dimensions, condition, likely performance, avoided carbon
  • Design under stock constraints: assign available elements, optimise, generate variants that fit the stock
  • Account for uncertainty: limited quantities, availability over time, technical and insurance qualification

Deliverables

  • A structured, reusable resource database
  • A script for design under stock constraints
  • A project sheet: reuse rate, avoided carbon, technical and insurance watch points

Audience and prerequisites

Architects, reuse consultants, PEMD diagnosticians, structural consultancies, clients, social landlords. Basic project knowledge; F5 useful for day 3 but not required.

SessionContent
Day 1 am · SearchFramework: the AGEC law, the PEMD diagnosis (mandatory since 1 July 2023 before significant demolition or renovation), resource diagnosis, REP PMCB; mapping stock sources
Day 1 pm · SearchAI-assisted watch and aggregation: automatic extraction of elements from PEMD PDF reports, geolocating stock, a resource map around the project
Day 2 am · DefineBuilding a resource database: a common data schema (type, dimensions, quantity, condition, location, availability date), classifications, links to FDES sheets and the INIES database
Day 2 pm · DefineAI characterisation: recognising elements in photos, measuring from point clouds or photogrammetry, estimating condition and residual performance, avoided-carbon calculation; limits and the need for testing
Day 3 am · DesignPredictive design under stock constraints: the assignment problem, combinatorial optimisation, variant generation in Grasshopper from the inventory (façade from reclaimed joinery, floor from recovered elements, mixed tile layouts)
Day 3 pm · DesignWorkshop on real, limited stock: each group designs a project element and measures reuse rate, avoided carbon and remaining stock; "partial stock" and "lost stock" scenarios; presentations

Methods: framing, demonstrations and workshops on a real dataset (PEMD report, photo inventory, 3D survey). Assessment: 40% resource database quality, 60% workshop project and presentation.

Practical details

How the courses work

100% online

Live virtual classes with the trainer and supervised hands-on workshops. A computer, a stable connection and a headset with a microphone are enough; the software needed is listed before the session.

Open or in-house

Open sessions on the published calendar, or in-house sessions for a firm, a local authority or a client organisation, on the date you choose and with an adapted case.

Assessment and certificate

Skills check before the course, assessment at the end, satisfaction surveys right after and at 3 months. Each participant receives a certificate of attendance.

Trainer

All courses are taught by Roland Dehghan Kamaragi.

Accessibility

A disability officer reviews each adjustment request before enrolment and arranges what is needed.

Pricing

For open sessions, a rate per day and per participant, depending on the technical level. For in-house sessions, a group day rate, on quotation.

FAQ

Your questions

Do I need to know how to code?

Not for most of them. F1, F2, F4, F6 and F8 require no programming skills. F5 requires experience with a 3D modeller (Rhino preferred) and F3 some knowledge of BIM.

Are the courses online?

Yes. All courses are 100% online, in a live virtual classroom with the trainer, with supervised hands-on workshops.

Can you run a session for our firm or local authority?

Yes. Each course can be run in-house for a firm, a local authority or a client organisation, on the date you choose and with a case adapted to your work.

How many participants per session?

Up to 12 participants for most courses, 15 for F2 and 20 for F6.

How much does it cost?

For open sessions, a rate per day and per participant, depending on the technical level of the course. For in-house sessions, a group day rate, on quotation.

Interested in a course?

Tell us the course, your preferred dates and the number of participants: we will reply with a programme and a quotation.

2IIA logo

Last updated: 19 November 2026