Vetle
Gangeskar

My own product. LURK collects advertising creative at scale, structures it, ranks it, and surfaces the patterns underneath — so a creative decision can start from evidence instead of a blank page.

I built it hands-on: Python and FastAPI for the services, PostgreSQL for the creative corpus, Redis and Celery for the scheduled jobs that keep it current without me touching it. Ingestion runs through a priority chain — the official Graph API first, falling back to lighter methods — so collection degrades gracefully instead of breaking when one route closes.

I use it as a second layer of creative judgment. Instead of starting every concept from nothing, I use it to see which patterns are running across a market, which creatives keep resurfacing, and where the unexplored angles are. It does not replace the judgment. It gives me a better starting point.

I own the creative cycle end to end: insight, hook and script, through production and editing direction, into performance analysis and the next iteration. Direct response for subscription products across four Nordic markets.

The volume is the point. Developing and testing 100+ creatives a month means taste alone stops scaling — you need named reasons a sequence worked, or you relearn the same lesson every week. So I work in failure modes: was it the first second, the pacing, the visual continuity, the clarity of the claim, or the offer? Each answer changes a different lever.

A large share of that output is now AI-generated or AI-assisted. I build the production workflows that raise video output, localization capacity and testing speed while holding product and brand consistency — and every AI-generated placement ships with an on-screen disclosure.

A standing part of the job is keeping current with the models themselves, especially video and image generation. New versions land constantly and each one changes what is worth attempting, so I test them against each other on real briefs — subject consistency, motion, artifacts, how far a prompt can be pushed before the model stops cooperating. Knowing which model to reach for has become as much of the craft as the edit.

And the week does not start on a blank page. It starts in LURK — my own product, open on a second screen, telling me what is already running in the category before I write a single line.

A Norwegian safety app that helps locate a buried person using GPS, Bluetooth and sound — built as a supplement to a transceiver, never a replacement for one. I work on growth and distribution.

Safety products have an unusual growth problem: the best product does not win automatically. It wins when it is the one people already trust — seen in the lift, in the ski bus, recommended by a guide, used by the people they ski with. That makes distribution a credibility problem before it is a media-buying problem.

So my job is the channel mix that gets a safety product in front of people before the season starts, and the measurement to know which of those channels actually moved anything. It is the same loop as the creative work — test, measure, move budget toward what holds — with a product where being right matters more than usual.