Cloudglue and Tinycloud
Cloudglue APIs and Tinycloud agents that turn video collections into searchable knowledge, with every answer traced back to the footage.
CTO
Kevin Dela RosaCTO, Cloudglue
I take AI to production.From research to real users.
I’ve spent 15 years shipping AI in consumer products and the infrastructure behind them. That’s product matching across Amazon’s catalog, visual search and OCR for hundreds of millions of Snapchatters, and now video agents at Cloudglue.
Open to speaking and consulting.
- publications
- 22
- patents
- 11
- users reached
- 100M+
Shipped at Super Bowl LV · Cannes Lions · Art Basel
Spoken at DEF CON · KubeCon · AWS re:Invent
Published at NeurIPS · ICCV · CVPR · WACV · KDD · ACM WWW · SIGIR · ACM RecSys · ACM CAIS · NAACL · ISMIR
Talks & demos
- 2026Evidence-Grounded Video Investigation over Large Video CorporaECCV Demos, Malmö
- 2026Overcast: Video OSINT Agent. Point It at 100 Videos, Ask AnythingDEF CON 34 Demo Labs
- 2025Autonomous Video Hunter: AI Agents for Real-Time OSINTDEF CON 33 Recon Village
- 2025AI That Gets the Picture: Building Video-Aware AssistantsCascadiaJS
- 2025Cloudglue: Turn Videos Into Structured, LLM-Ready DataGen AI Zoo, Seattle Tech Week
- 2020MLOps at Snapchat: Continuous Machine Learning with Kubeflow and SpinnakerKubeCon + CloudNativeCon NA
- 2019Build Accurate Training Datasets with Amazon SageMaker Ground TruthAWS re:Invent
Research
- 2026Tool-Skills vs. Workflow-Skills: Routing and Composition in a Multimodal Agent HarnessACM CAIS AgentSkills Workshop
- 2025Smart Routing for Multimodal Video Retrieval: When to Search WhatICCV Multimodal Representation and Retrieval Workshop
- 2025VideoMCP: Video Intelligence for Multimodal Agent ReasoningKDD Agentic AI for Enterprise Workshop
- 2025RAVEN: An Agentic Framework for Multimodal Entity Discovery from Large-Scale Video CollectionsAgent4IR at AAAI
- 2025Video Enriched Retrieval Augmented Generation Using Aligned Video CaptionsSIGIR MRR Workshop 2024 · WWW Companion
- 2023FORB: A Flat Object Retrieval Benchmark for Universal Image EmbeddingNeurIPS Datasets and Benchmarks Track
- 2021Open-Vocabulary Object Detection Using CaptionsCVPR
Also an inventor on 11 patents in visual search, scanning, OCR, and ML systems.
Work
Products, open-source tools, and systems shipped at scale.
Overcast
A video OSINT agent that searches, listens, matches, maps, and remembers across a collection of video.
Creator
mDex
An offline card scanner and open-source visual matcher. A computer-vision rabbit hole, built for a real card show.
CreatorMore work
- Autonomous Video Hunter
A video-agent system for real-time OSINT: find targets, extract signals, and generate grounded reports.
- Snapchat Scan and Visual Search
Computer vision and retrieval systems shipped into Snapchat camera, Scan, search, messaging, and AR.
- ML Infrastructure at Snap
Continuous ML, labeling, training, deployment, and GPU infrastructure for camera intelligence and gen AI.
- Applied AI, Search, and Language Tech
Product matching and de-duplication across Amazon’s catalog, plus NLP research at Carnegie Mellon.
Writing
I put every CSI button in a CLI
Building Overcast: the real algorithms behind the hacker-movie tropes, connected to an agent that remembers.
I vibe coded an offline card scanner for the National
10,023 trading cards, an offline iPhone scanner, and a return to classic computer vision.
How to grep video
Why transcripts aren't enough, and how structured representations make video searchable by agents.
Speaking & consulting
I give talks and live demos on video agents, retrieval, and production ML, and consult on taking AI systems to production. Connect with me on LinkedIn to get in touch.
