About
Research to real-world systems.
I’m Kevin Dela Rosa, co-founder and CTO at Cloudglue. My work connects machine learning research, product development, and the infrastructure needed to ship both.
Before Cloudglue, I spent eight years at Snap working on computer vision, visual search, and ML infrastructure. Earlier chapters include Amazon, language technologies at Carnegie Mellon, and software engineering and physics at UT Arlington.
multimodal AI, retrieval, NLP, vision, and learning systems
11patentsvisual search, scanning, AR unlocks, OCR, and ML systems
100M+users reachedSnap camera, Scan, search, messaging, and AR experiences
50+ML use casespowered by vector retrieval, visual search, and data pipelines
0 to 10team buildgrew and led a production ML infrastructure team at Snap
The longer story
Snap Inc.
Staff Machine Learning Engineer and Tech Lead Manager
- Shipped computer vision, search, camera, messaging, and Scan systems to iOS and Android at Snapchat scale.
- Built large-scale vector search, OCR, ML data systems, continuous training, and generative AI infrastructure.
Amazon
Software Development Engineer I to II
- Built product deduplication systems that improved Amazon catalog quality across 30+ product categories.
- Built workflow orchestration for massively scaled, mission-critical distributed applications.
Carnegie Mellon University
M.S. Computer Science, Language Technologies
- Focused on applied machine learning, information retrieval, language technologies, and multimedia analysis.
- Graduate researcher in the Language Technologies Institute on the REAP intelligent tutoring system, running studies with ESL learners on instruction methods.
University of Texas at Arlington
B.S. Software Engineering and Physics, Summa Cum Laude
- Dual focus on software engineering and physics, graduating Summa Cum Laude.
- Research in the Vision-Learning-Mining lab (recommendation ranking, boosting, query-by-humming) and computational astrophysics on solar supergranule rotation using NASA SOHO/MDI data.
Early research & internships
Space and Naval Warfare Systems Center Pacific
Naval Research Enterprise Intern
Built text classifiers (Naive Bayes, Rocchio, k-NN, and SVM) to categorize micro-text in military chat — work that led to an ICMLA 2009 publication.
National Institute of Standards and Technology
Summer Undergraduate Research Fellow
Developed an AJAX interface for FireEye, a software testing tool, as part of a research project on automated combinatorial software testing.
Teams & institutions
Venues & stages
Publications, presentations, and places where the shipped work has appeared.
Published in
NeurIPS · ICCV · CVPR · WACV · KDD · ACM WWW · SIGIR · ACM RecSys · ACM CAIS · NAACL · ISMIR
Browse the publicationsOutside the day job
I collect trading cards and build things for the problems I run into. That’s how a card-show trip turned into mDex, an offline scanner and a fresh look at a research benchmark.