WES
2026A research data platform for daylight research at TU Berlin
Design and prototype of a data platform for the Chair of Lighting Technology at TU Berlin, bringing highly heterogeneous daylight measurements into one queryable model. The underlying research uses daylight simulations in VR to treat symptoms of daylight deprivation, and it depends on data that currently lives in different formats, in different places, and partly in individual researchers' heads.
The Challenge
The data arrives in four kinds, each with its own format, resolution and naming: spectral readings from a rooftop measurement site recorded over years, HDR and fisheye photography, questionnaires on subjective perception, and atmospheric and weather records. Against TU Berlin's research data guidelines the FAIR principles are only partly met: much of it is not centrally findable, and access can depend on knowing who to ask. Media files add a second problem, with images in the tens of megabytes and video running to several gigabytes, so an interrupted upload cannot mean starting over.
The Solution
A relational database holds structured metadata, measurement setups and survey data alongside references to media kept in object storage, so large files never travel through the database. A single API carries authentication, authorisation and resumable transfer for the large files. Between the sources sits a translation layer built on work that already exists rather than a new standard: the SKYSPECTRA data package from CIE TC 3-60, and the chair's own KEY schema, which identifies a measurement series by location, timestamp and setup. Formats, naming and time references are mapped into one model researchers can filter and compare across.