Gabriel Parra

I’m a Machine Learning Engineer with 8+ years of experience leading the design, development, and deployment of deep learning models from concept to production for startups and enterprise teams. I’m a hands-on leader who drives end-to-end delivery—designing and training models, running large-scale offline experimentation, validating impact with rigorous A/B testing, and ensuring reliability post-launch.

At Meta Platforms Inc., I helped lift ad revenue by $150M+ annually through CTR/CVR and negative feedback prediction model launches. I led a privacy migration across 100+ ads ranking and quality neural network models serving billions of impressions, and I preserved performance by replacing restricted features with privacy-safe alternatives and re-optimizing feature sets under no-regression constraints using A/B testing.

I also eliminated 10K+ engineering hours and 27M GPU-hours from experimentation by designing and introducing an ML experimentation platform with a centralized UI, automated experiment scheduler, and a declarative framework to scale offline experimentation for 500+ new ad ranking features. Earlier, at AutoFact, I oversaw pricing models that improved accuracy by 20%, built computer vision pipelines for structured vehicle attributes, consolidated analytics data on AWS, and reduced preprocessing time by 80%+—enabling faster, reliable ML retraining at scale.