AI Matchmaking

AI Matchmaking is the headline feature of the admin SaaS. It surfaces the three to seven cities most useful to you — peers with similar challenges, complementary strengths, or natural twinning fit.

How it works #

1. Vectorisation. Each city's profile (Pulse sub-scores, demographics, climate zone, economic mix) is encoded as a 256-dimensional vector.
2. Distance computation. Cosine similarity between your city and every other city in the corpus.
3. Re-ranking. A small instruction-tuned model re-ranks candidates against your stated matchmaking intent (peer, complement, twin, sister).
4. Explanation. Each match comes with a one-paragraph "why this match" written from the underlying indicators.

Match types #

Peer

Cities of similar size, climate, and Pulse band — useful for benchmarking.

Complement

Cities strong where you are weak (and vice versa) — useful for staff exchanges.

Twin

Cities with formal twinning agreements with you — see Sister Cities Hub.

Aspirational

Cities one Pulse band above you, with similar starting conditions a decade ago.

Example: İstanbul looking for transport peers #

İstanbul's transport pillar is 71/100 with growing demand. The top three peer matches surfaced by AI Matchmaking in April 2026:

• Seoul (transport 88) — aspirational; metro coverage at İstanbul-scale population.
• Mexico City (transport 64) — peer; comparable polycentric structure and BRT investment.
• Warszawa (transport 76) — twin; existing sister-city agreement, joint transport workshop already booked.

---