MOOD FILTER
HP x AKQA
A proof of concept retail installation for HP's "Made to Be Less Hated" Smart Tank printer campaign — an in-store AR face filter that transforms passers-by into exaggerated sad or happy expressions, toggling in sync with rotating headlines that pair printer frustrations with HP features. Eduardo led the technical direction and built the PoC for deployment in Elkjøp and Currys stores.

OVERVIEW _
The HP Smart Tank campaign was built around a simple, honest insight: people hate printers. The Mood Filter installation leaned into it directly. As a customer passed an end cap display, proximity detection would wake the screen; a real-time AR face filter would pick up their face via webcam and map it to an exaggerated sad expression. Every 15 seconds, the filter and headline rotated — the sad face giving way to a happy one, as each printer frustration was answered by an HP Smart Tank feature (ink cost, WiFi reliability, eco credentials, setup simplicity).
Eduardo led the technical direction and developed the proof of concept, designing the face filter pipeline, proximity-triggered sleep/active mode, and the headline-filter synchronisation layer. The installation was designed for end cap deployment in Elkjøp stores across Denmark and the Nordics, and Currys stores in the UK — two distinct hardware footprints, both explored in the brief.
The concept was well received by HP but ultimately shelved due to GDPR constraints on facial recognition in European retail environments.
TECHNOLOGY _
AR face filter (real-time, webcam-based), proximity detection, PC with NVIDIA GPU, non-touch display (42–55", portrait), custom UI (headline and filter synchronisation)
RELATED _

Global Platform
Headless microservice platform across 40+ markets — 57% uplift in users, one-click market launch via environment replication.

Digital Memories
Cloud-connected museum experience bridging RFID wristbands to a personal media retrieval portal — first-party data capture from 2% to 39%.

Museum AI Platform
AI-powered archive — 114,000+ records from 7 legacy systems searchable via natural language, results enriched by vision models.