How it works
Dundir is being designed to connect mathematical procurement optimization with physical verification at the gate and an accumulated contract memory. The professional evaluates and decides. The system is designed to compute, verify, and archive.
The lifecycle
The design in five steps, from specification to handover.
- 01Specificationdesign
- 02Solverresearch and design
- 03Gate checklab prototype
- 04Wkb dossierdesign
- 05Contract memorydesign
Structuring the specification
In the design, an AI language layer reads your specification or tender document in plain language and extracts hard constraints, delivery schedules, budget ceilings, and CPR 2024 certification criteria. This is meant to be the only stage where a language model operates.
Deterministic solver calculates the optimum
The optimization engine is to weigh unit price, delivery schedules, freight miles, and carbon intensity. Pure mathematics by design: identical inputs must yield identical outputs every time, with an audit-ready reasoning trail. The solver is in research and design.
Smart gate verification and 3-way matching
The goal: at the site gate, mobile or fixed edge OCR reads delivery tickets and truck license plates, and the system matches purchase order, delivery receipt, and vendor invoice. Today the first edge hardware runs on the bench, without a camera yet. OCR and matching come next.
Real-time Wkb handover dossier compilation
Inspection records, delivery receipts, and quality stamps are to be linked automatically into the statutory Dossier Bevoegd Gezag under the Dutch Wkb regulation, so that no documentation goes missing at final municipal commissioning.
Feedback to accumulated contract memory
Actual supplier delivery punctuality, defect frequencies, and unit price variances are to be logged automatically and to recalibrate the scoring parameters for the next tender.
Data sovereignty
Why it is designed to run on your own infrastructure.
Friðriksson 2025, KTH · Van Duuren 2025, TU Delft · BIO 2019
Explainability
Every calculation and gate inspection must be verifiable.
Deterministic
Identical inputs yield identical outputs. Given identical inputs, generic enterprise AI moved one subcontractor total by up to 8,354 euro between sessions.
Hamppi 2025, Aalto University · Atil et al. 2025
Which constraints were enforced, what trade-offs were balanced between freight transit distance and material unit costs, and why a supplier was selected: the system is designed to record the exact formula and parameter weights.
The same rigor applies at the gate: every delivery ticket scan and 3-way match decision is to be logged with timestamps, reference IDs, and variance percentages.
For municipalities, this is meant to support the reasoning required under Public Procurement Act Article 2.130. For contractors, it is to provide a documented file during subcontractor disputes or claims.
Boundaries
What the system does not do.
Dundir does not make autonomous business decisions and does not replace human experts. The estimator, buyer, and site superintendent maintain full control.
The platform is meant to take over clerical data entry and manual matching so professionals can focus on negotiations, quality inspections, and strategic vendor management. Formal sign-off and legal accountability remain with your organization.
The system is to calibrate against your own transaction history, so that accumulated vendor intelligence stays within your firm when experienced staff leave.
Four ways in
Where you come in depends on what you need.
Become a research partner
Construction SMEs join the research track with Windesheim. No cost, and a say in what gets built.
What it involves Way in 02Pilot as a design partner
The system is not finished. You decide what it gets calibrated on, and you own the result.
Terms Way in 03Procurement scan
Ten questions about your current process. Report on screen straight away. No email address needed to read it.
Start the scan Way in 04A 30 minute conversation
No deck, no pitch. We ask the questions.
Request a call