Skip to main content
← engineering

Cloudglue Expands Multimodal Video Infrastructure Team After YC S24

By Zero G Talent•

Video represents a massive share of the internet's data, but most generative AI systems still treat visual and audio streams as simple text transcripts. San Francisco-based Cloudglue, founded in 2024 out of Y Combinator's S24 batch, builds developer APIs and workflow agents designed to let large language models process, index, and reason across raw video and audio feeds.

Operating with a core team of three employees, Cloudglue is now recruiting its first wave of technical hires to expand both its developer-facing platform and underlying multimodal models. The company has opened three engineering positions across infrastructure, full-stack systems, and research engineering, each offering compensation between $120,000 and $250,000 with remote flexibility across San Francisco, Seattle, and Los Angeles.

Developer APIs for Video Intelligence

Cloudglue provides developer infrastructure centered on three primary capabilities: extracting structured data from visual feeds, generating play-by-play multimodal descriptions, and querying collections of video through chat completion APIs.

In a presentation at Gen AI Zoo - Seattle, the team demonstrated how these building blocks operate in practice. Developers can send a video URL through the extract API and specify a structured schema—such as pulling out ingredients, recipe names, and sequential steps from a cooking broadcast—to populate a web application directly. The describe capability produces detailed breakdowns of scenes across audio and visual modalities, integrating with standards like the Model Context Protocol (MCP) to plug into external tools and create multi-step summaries like illustrated storyboards. The system also supports chat completion, letting users query video archives to isolate specific visual features, such as locating 3D animations within conference talks.

Beyond headless APIs, the startup introduced Tinycloud, an AI agent for video content workflows. In a product demonstration, CEO Amy Xiao showed the agent watching episodic footage to create clip galleries, burning synchronized captions into video for publishing, synthesizing multi-episode podcasts into written posts with source clips, and pulling travel destinations from short-form clips onto a map.

Technical Leadership

Cloudglue was founded by Amy Xiao and Edward Zhou.

Amy Xiao serves as Chief Executive Officer. Her background in machine learning and infrastructure spans early-stage research to enterprise production. She entered the field in 2016 as one of the first 10 ML research engineers at Borealis AI, later joining Snapchat's Perception team where she built ML infrastructure that powered Scan, the platform's real-time visual search product. Before co-founding Cloudglue, Xiao worked at AWS and served as an ML solutions architect at Arize AI, consulting with engineering teams at Fortune 200 enterprises on observing and monitoring production machine learning and LLM pipelines.

Scaling the Founding Engineering Team

Cloudglue's current openings cover the full cycle of multimodal engineering, from novel model research to async backend pipelines and developer-facing surfaces.

The Research Engineer position focuses on designing core multimodal retrieval and video reasoning systems to extract structured data from video content. The role spans research and production engineering, tasking the hire with taking novel approaches to retrieval and understanding from prototype to final implementation.

To manage the high compute and throughput requirements of video workloads, the Founding Engineer, Infrastructure will design and scale backend systems for Cloudglue's video processing pipelines, search and retrieval infrastructure, and asynchronous job orchestration. The engineer will build distributed systems, manage ML inference serving, and operate indexing infrastructure that converts raw footage into searchable data.

On the product layer, the Founding Engineer, Full Stack will build and ship features across Cloudglue's APIs, developer dashboard, SDKs, and technical documentation. This role links the platform's machine learning outputs—such as transcriptions and visual analyses—into usable software surfaces and developer tools.

Open Positions at Cloudglue

All three engineering roles are open to candidates in San Francisco, Seattle, and Los Angeles, with remote flexibility within those hubs.

Role Location Work Mode Salary
Research Engineer San Francisco, CA / Seattle, WA / Los Angeles, CA Remote $120,000 – $250,000
Founding Engineer, Infrastructure San Francisco, CA / Seattle, WA / Los Angeles, CA Remote $120,000 – $250,000
Founding Engineer, Full Stack San Francisco, CA / Seattle, WA / Los Angeles, CA Remote $120,000 – $250,000

See all open roles at Cloudglue