Most governments run on hindsight. A road cracks, then someone fixes it. A tube delays, then commuters scramble to find another route.

Two Imperial College London students think governments can do better.

⁠Dhruv Syam⁠ and ⁠Viktor Smirnov⁠ are building Arakon – a predictive intelligence layer for government that turns city data into a digital twin, allowing policymakers to test decisions before implementing them.

The proof? They’ve segmented Hammersmith & Fulham into 17,000 road chunks, each scored for the likelihood and severity of future damage. That helps the council prioritise maintenance before small issues become costly disruptions.

The same engine can scale beyond roads to bigger targets: tube closures, traffic jams, surge pricing. Transport for London might want to take a look.

The clever part is that they don’t do it by surveilling our data. The raw data never leaves the government’s servers. Models are trained locally without sharing it out. That’s federated learning, the technology FLock is providing.

What starts as a pilot in Hammersmith could become a global public good. Built in the UK, not the Bay, it’s another piece of the bigger picture behind the UK’s sovereign AI ambitions.

Timestamps:

00:00 – 03:45 | Meet Dhruv Syam & Viktor Smirnov: Reimagining government with predictive AI and city-scale digital twins.

03:45 – 09:40 | Why cities should be modelled as connected systems: Road maintenance, NHS pressure, and cross-domain signals reveal the second- and third-order effects of policy decisions.

09:40 – 14:43 | From potholes to predictions: How AI forecasts road deterioration before it happens.

14:43 – 18:40 | Beyond roads: Predicting traffic, tube closures, and city-wide disruption with the same AI engine while keeping sensitive government data secure.

18:40 – 32:10 | The case for sovereign AI in the UK, how federated learning enables it, and what makes FLock the right infrastructure for privacy-preserving public-sector AI.