Hans Del Mundo

delmundo.hansmatthew@gmail.com GitHub LinkedIn ORCID

Aspiring computer science person.

I am passionate about computing and what we can achieve with these wonderful machines. My interests now focus on representations and interpretability, and more broadly on AI integration and alignment.

Not does it work? but how and why it works.

Publication

A Graph-Based Approach for Efficient, Low-Transfer Routing in Flexible Urban Jeepney Systems

Lead author · Journal of Transport Geography · 2026

Jeepney routes have no fixed stops or timetables, which breaks the assumptions behind ordinary transit graphs. TFTFGraph is a novel representation of jeepney structure: each route is a single node, and coordinate-level transfer edges carry distance and fare. On Cagayan de Oro data the network shrinks from 65,244 nodes to 82 nodes.1 Routing finds trips with fewer unnecessary transfers than Dijkstra and A*,2 and queries stay fast.3

65,244 nodes
82 nodes
Graph size before and after spatial hashing. At true scale the 82-node bar would be under half a pixel wide, so it is drawn at a minimum width.