ΕΡΓΟΝ Γ΄
Urban IntelligenceResearchCausalCityAI
“Cities behave less like maps and more like living systems.”
CausalCity builds imaginary cities and simulates their traffic with real physics, so every traffic jam it produces comes with a known, labeled cause. It's built to become the foundation of a platform that could one day forecast what a city's traffic will do, test changes against it, and recommend a decision.
Timeline
Loading diagram…
Ι — Origin
How it started.
CausalCityAI began during a Google hackathon, first as a challenge problem, eventually as an obsession. We got fascinated by the idea that locals understand a city in ways software never does. People know the patterns, the hidden behaviors, the traffic habits, the events, the rhythm of a place. What if a system could learn that too?
ΙΙ — The Problem
What we're solving.
Navigation apps tell you what traffic is doing right now. They don't tell you why, and they definitely don't tell you what happens if you reroute five thousand cars down a side street. Municipalities spend millions on infrastructure without knowing whether it will trigger the Braess Paradox, the effect where adding a road mathematically makes traffic worse.
Cities are dynamic systems. Everything in one loops back on everything else: traffic shapes behavior, behavior reshapes traffic, weather sits underneath both. Pull one thread and the rest moves. A model that tries to predict the future, explain why it happened, and recommend a route all at once will be bad at all three.
ΣΦΑΛΜΑ — What We Got Wrong
We designed the six layers assuming we'd build through them in order. In practice, the layers that reason about a city, forecasting, causal discovery, decisions, only mean something once they run against real data, and real sensor data isn't in place yet. Building them now would mean testing against numbers we made up ourselves. We stopped at the data layer on purpose instead of shipping placeholder intelligence to round out the pitch.
ΙΙΙ — Where It Stands
Told straight, no jargon.
Right now, CausalCity is a very good traffic generator, not yet a decision platform. It builds twelve imaginary cities with realistic road networks and lets you watch traffic build, jam, and clear, with the cause of every jam known and labeled, because we generated it.
That sounds small, but it solves a real problem. Researchers who want to teach a model to find the cause of traffic, not just predict it, can't get real cities to hand them labeled cause-and-effect data. Real sensors don't come with ground truth attached. Ours does.
The bigger pitch, forecast what a city will do, simulate what happens if you reroute it, recommend a decision, is the destination, not where we are today. The diagram below shows both: what's running now, and what's still just a target architecture.
Want the wiring diagrams, the real stack, and what’s still 0% built?
IV — Today
What it can already do.
- ›Twelve archetype cities stitched into one continuous, physically simulated road network
- ›Traffic physics accurate enough to reproduce the counter-intuitive cases, like a new road making traffic worse, not just ordinary congestion
- ›180 days of causally-linked synthetic data generated in under 30 minutes
- ›A live, interactive map of the whole simulation, not static charts
V — Who It's For
Who we're building for.
AI/ML Researchers
"Generate 180 days of causally-labeled synthetic traffic data in under 30 minutes. You know the exact cause of every anomaly, because we generated it."
Municipal Governments & Urban Planners
"This is what we're building toward next: a counterfactual simulator you could test infrastructure changes against before you pour concrete. Would a municipality actually reach for this, or are we solving a problem nobody has budget for?"
ΜΕΣΑ — Media
CausalCityAI walkthrough, the generator and viz, end to end
VII — Open Questions
What we don't know yet.
- Should we ship cityviz and the dataset as a standalone tool now, before building the intelligence layers?
- Can synthetic data replace real sensor deployments for model training?
- What city-level decisions benefit most from a "what if we changed this" testing tool, once it exists?
- How clearly does the reasoning need to be explained before a city government would actually trust and adopt it?
Feedback wanted from: Urban planners, transportation researchers, mobility startups, governments, traffic engineers, systems engineers, and people who think this idea cannot work.
We’re not asking you to validate this. We’re asking you to break it.
ΟΜΑΔΑ — Team
Nikhil Y N
CEO & Technical Lead
Prithvi K M
Engineer, Cloud Infrastructure & Python
Karan S J
Engineer, Web Development
M. Talha
Engineer, Database Architecture & Data Pipelines
Anoushka P
Engineer, AI/ML
Ahad U B
Engineer, AI/ML