Ricardo de Deijn · Applied AI Engineer
← Portfolio

Highlights

The longer version.

Research that made it out of the lab, hackathons decided in a weekend, and eight talks spent making hard architectures legible to people meeting them for the first time.

Synthetic image generation research code

01 / Research

Research & development

My research sits where computer vision meets scarce data: real-time snow detection for pedestrian safety, and synthetic data generation for the cases where the real images simply do not exist in enough quantity to train on.

That work has been published at MWAIS, AIMLA, and CADSCOM, where the paper on prompt-based generated images won Best Paper and was fast-tracked for journal publication in JMWAIS.

Current interests run toward federated learning for mobile AI and spatial attention mechanisms inside deep architectures. Both aim at the same question of how models behave once they leave a controlled environment.

Data Derby hackathon team

02 / Competition

Competitive data science

I compete in data science hackathons and have placed near the top at Data Derby and MUDAC. Predictive modelling, time series, and visualisation, all under a clock with a team you met that morning.

The technical skills matter less than the constraint. A weekend forces you to decide what is actually worth modelling, which is the same judgment that production work demands, just compressed.

Ricardo de Deijn presenting at a conference

03 / Speaking

Speaking engagements

I have presented at MWAIS, AIMLA, and CADSCOM, three times for MinneAnalytics at Best Buy HQ in Richfield, and twice as a guest lecturer for the DREAM student organisation. The recurring brief is the same: take an architecture that sounds impenetrable and hand an audience a working mental model of it in forty minutes.

Talk archive