{Tech:Europe} Stockholm Hackathon

Yongkang Zou · 2026-04-09 · hackathon · Spatial Intelligence

Dianoia (DNO1) - AI crime investigation system

One week after arriving in Stockholm, I finished my first local hackathon and placed 3rd as a solo team.

For this event, I built Dianoia (DNO1), an open-source Palantir for AI crime investigation. It reconstructs 3D scenes, reasons through the chain of evidence to infer what happened, and provides an interactive suspect profiling system that can be continuously refined.

Presenting the Dianoia UI

The platform relies on several core features to process and analyze crime scene data. It includes a 3D blueprint viewer alongside a 2D SVG floor plan view. The Marble 3D reconstruction pipeline handles the spatial mapping of crime scenes.

Demonstrating the 3D reconstruction pipeline

Investigators can use evidence placement tools equipped with type badges and credibility scoring. The backend uses a Gemini reasoning pipeline with progressive S0, S1, and S2 stages for hypothesis generation. This connects to a timeline playback feature with actor interpolation and an evidence relationship graph. Finally, the NanoBanana suspect profiling module helps track and evaluate potential actors.

Showing the suspect profiling feature

The system is built with a Go backend and a TypeScript frontend. You can view the source code on GitHub.

Dianoia project demo animation

A huge shoutout to Bela Wiertz, Amit Vadi, and Sophia Nabil Gustafsson. Thanks to {Tech: Europe} for organizing this event and for the amazing venue at Lovable HQ, with sponsorship from Google DeepMind and Dust.

Photos