Jurian Baas

Creative technologist at the City of Amsterdam's AI Lab. I research, design and build AI-enriched products end-to-end — from user interviews to design to code to deployment.

At the AI Lab I build high fidelity AI prototypes to find out how AI can help citizens and public servants, and what role the City should give it. On these projects I lead the design and frontend; the LLM pipelines and RAG are built with an AI specialist on the team (engineering I do myself on smaller projects). I also lead the AI and Design subgroup of the City's design guild, where we think about what AI means for our craft, and what it means to design for a world where AI keeps getting more influential, for good and for bad. Before the AI Lab I was a technologist at a multidisciplinary design research team working on subjects like affordable housing, the energy transition, and accessibility.

I have a background as a political scientist, web designer, copywriter, UX researcher, and data visualization developer.

Email me at jur@jurian.me.

Selected projects

Subsidiemaatje prototype screenshot

Subsidiemaatje

AI prototype for the City of Amsterdam

  • AI prototype helping Amsterdammers find, choose, and understand the city's ~100 subsidy schemes, built at the Amsterdam AI Lab in Q4 2025 in opdracht van directie SIS (Subsidies, Inkoop, Juridisch Bureau Sociaal).
  • Preceded by a three-month discovery phase: stakeholder interviews, a working RAG-based question-answering tool over the existing subsidy texts, and a Figma prototype tested with users, all used to scope what the build phase should become.
  • Designed and built the Next.js frontend on the Amsterdam Design System: AI-element UI language (purple highlight, transparency notices) and the parallel “personal explanation / official source text” pattern that always keeps the legal regulation one tab away.
  • Wrote a custom amsterdam.nl scraper after the site stopped being scrapable. It's now whitelisted by the city's web team and feeds a RAG layer with up-to-date regulation text for all 100+ schemes.
  • Co-built with an AI specialist: the Flask backend, the multi-step LLM pipelines, and the synthetic-recall eval (5 fictive cases per regulation, measured across 99 schemes).
  • Tested with 10 Amsterdammers; most found subsidies faster than via Google or amsterdam.nl and felt more confident in their choice. Code open-sourced; classified as a low-risk AI system by the city's Digital Strategy directorate.
Inkoop Schrijfhulp prototype screenshot

Inkoop Schrijfhulp

AI prototype for the City of Amsterdam

  • AI authoring environment helping Amsterdam's procurement advisors write inkoopstrategieën, built at the Amsterdam AI Lab in Q4 2025, same client as Subsidiemaatje.
  • Architected the three-level prompt instruction hierarchy (project / document / chapter) that lets the same engine support different writing tasks, teams, and document types.
  • Designed and built the Next.js frontend (chapter editor with auto-save and version history; human-validation checklist gating chapter approval; admin/template UI), and expanded the FastAPI backend and PostgreSQL schema to fit the hierarchy.
  • Co-designed prompts with five procurement advisors across three 90-minute sessions; testers self-reported time savings of 50% to 2–3 working days per inkoopstrategie, and 100% wanted to keep using the tool.
  • Co-built with an AI specialist: the Azure OpenAI integration and the autonomous web-search agent (browser-use + Playwright). Code open-sourced.
AI Futures prototype screenshot

AI Futures

Speculative AI research for the City of Amsterdam

  • Speculative AI research at the Amsterdam AI Lab in 2025, in opdracht van directie Dienstverlening, exploring two scenarios for the future of municipal services: AI quietly supporting civil servants behind the scenes, and citizens delegating tasks to their own AI agent that negotiates with the city on their behalf.
  • Co-designed the prototypes and built the SvelteKit frontend, including an i18n layer that detects the spoken language and switches the UI to match. You talk to the city in your own language and the interface follows.
  • Co-built with AI specialists and another designer: the LLM pipelines (Azure OpenAI, GPT-4.1) backing the two scenarios end-to-end.
  • Concrete use cases tested: contesting a parking fine and applying for a neighborhood event subsidy.
  • Demoed at the World Summit AI by Innovation lead AI Swaan Dekkers.
AI and Design blog post header

AI and Design talk

Lead of the AI & Design chapter of the design guild

  • I lead the AI and Design subgroup of the City of Amsterdam's design guild, helping designers build shared understanding of AI in their practice.
  • Kicked off with a presentation on what AI means for design work, published as a blog post.
  • Follow-up sessions so far: a group session with Wietse Hage, and a workshop on ethical considerations of AI in design.
  • Currently writing design principles for two contexts: building products that contain AI, and living in a society where AI is gaining influence.
Brainstormhulp ideation tool screenshot

Brainstormhulp

AI ideation tool for the City of Amsterdam

  • Ideation tool that brainstorms ideas from How Might We questions, using quotes from user interviews as extra context for the model.
  • Lets you pick a form to brainstorm in and see which quotes corroborate each idea.
  • Designed and built end-to-end; used in an ideation session with policymakers.
Ontwerpen van Vertrouwen workshop screenshot

Ontwerpen van Vertrouwen

Workshop tool for the City of Amsterdam

  • Interactive workshop helping civil servants reflect on citizen-government interactions using the YUTPA framework, a multi-faceted model of trust. In raw form, the framework demands too much of facilitators and participants for daily use.
  • Designed and built an accessible workshop format that digitally guides facilitators and participants through every step. People weigh in on the case individually beforehand; the on-screen flow then walks the group through structured discussion and into concrete next actions, with printed materials in support.
  • The workshop aims to make citizen-government communication more conscious and contribute to mutual trust.
Generative AI in policy slide deck

Generative AI and policy

Opinion piece with working prototypes

  • Opinion piece on how generative AI can both help and harm policy work.
  • Concrete examples: a 'document thinking help' (helpful, supports the writer's reasoning), a 'document fixer' (less helpful, replaces it), and an autogenerator of questions for the city council (damaging).
  • Most examples come with a working prototype, so the arguments are about tools you can actually try.
PNI software design document

Software for Participative Narrative Inquiry

Software design document

  • Participative Narrative Inquiry (PNI) is a research method by Cynthia Kurtz that uses experience stories to make sense of complex situations. Participants share stories and answer questions about their own and others' stories; stakeholders help surface the patterns.
  • Wrote a detailed design document for software to support the method: drew the UI, mapped costs, compared vendors, and laid out what it would take to build it from scratch.
  • Goal: give the team a clear basis for deciding whether to buy, build, or assemble from existing tools.
Credit-trading game screenshot

Mobility credit-trading game

Live game for a TU Delft mobility research group

  • Live, Kahoot-style game built around a dynamic credit-trading mechanic, made for a TU Delft mobility research group. Players joined from their phones while realtime trading was projected on the room's screens.
  • Designed and built end-to-end (SvelteKit frontend, PocketBase backend, realtime event handling), deployed on RND servers.
  • Played at 'Mobiliteit als schaars goed' at Pakhuis de Zwijger by a room of about 50 participants.
Haltebuddy mobile prototype screenshot

Haltebuddy

Prototype, shipped into the GVB app

  • Before we built the prototype we interviewed many people by showing them Figma screens and an Observable notebook. We also designed and tested the scoring function of the public transport stop completely in the notebook. This separation of concerns is a keeper!
  • Prototype to prove information on public transport can be personalized according to the needs of people with a motor disability.
  • The prototype works with real, current data and is designed to be accessed as a web application on a mobile phone. Made with Vue and Vuetify.
  • Made together with Cliëntenbelang Amsterdam, GVB, and VRA.
  • We also made an observable notebook with interactive documentation for the team that is going to adapt parts of the prototype.
  • GVB integrated the functionality into their app as 'Haltehulp' after two code review sessions in which their developers (Infi) rewrote the prototype code for production, the kind of maker-to-maker handover I had hoped for. Presented at Inclusive Mobility Lab 2021 to push for national adoption.
Dashboard for unnamed client

Interconnected graphs and maps

Dashboard

  • This client had a large collection of complex, interconnected data. We made a bespoke dashboard with connected maps and charts to provide valuable insight into the data.
  • Used Svelte, D3 and Tailwind together for the first time, can recommend!
Home owners project tool

Project tool for home owners

Prototype

  • Organizational tool for home owners that would like to sustainably renovate their homes with the home owners association.
  • Personalized information for different types of associations.
  • Remembers the steps you took and allows for notes.
  • Simple CMS with Airtable.
  • Active on the Nieuw Amsterdams Klimaat website, ready for version 2.
Buurtcompost in Amsterdam thumbnail

Neighborhood compost in Amsterdam

Hobby project

  • Inspired by an interview with a data analyst working at the municipality of Amsterdam.
  • Script calculates the walking distance to the nearest wormhotel for all addresses in Amsterdam.
  • This way we can see where it might be handy to place more wormhotels.
  • Uses the routing engine OSRM and Open Route Service.
  • Can be used for other objects in Amsterdam, like pingpong tables, trash containers, pools, etc.
Social & Emotional Skills thumbnail

Social & Emotional Skills

The International Rescue Committee

  • Analysis and visualization of qualitative data gathered by the International Rescue Committee in the refugee camp Nyarugusu, in Tanzania.
  • Built a data pipeline with Python / Pandas to calculate ratios and percentages.
  • Made visualizations with D3.js and Billboard.js.
  • Made a presentation template in HTML, CSS, and Python.
Network visualization thumbnail

Interactive network graph

Prototype

  • Network graph showing average sentiment between characters in a TV show, derived from their dialogue across a chosen date range.
  • Made with D3.js and JavaScript.
The Internet Health Report thumbnail

The Internet Health Report

The Mozilla Foundation

  • Researched, collected and structured the underlying dataset for the first installment of Mozilla's annual Internet Health Report.
  • A design agency took the dataset through to the published report.