TwelveLabs

Video has quietly become one of the largest untapped data sources inside modern enterprises. Every broadcast, security recording, customer interaction, training session, product demonstration, and surveillance feed captures information that most organizations still struggle to search, analyze, and reuse. TwelveLabs was founded to make that video intelligence accessible.

Headquartered in San Francisco with a significant presence in Seoul, TwelveLabs is a video-native AI company co-founded by Jae Lee, Aiden Lee, Dave Chung, and Soyoung Lee. Rather than extending text-first language models into video, the company has built foundation models designed to understand visual scenes, audio, motion, and context over time. That distinction matters because enterprise AI is expanding beyond documents and chatbots into perception-rich workflows.

About TwelveLabs

Founded in 2021, TwelveLabs focuses on helping machines understand video through multimodal AI. Its platform enables organizations to search, summarize, analyze, and reason across video content using natural language instead of relying on timestamps, manually applied metadata, or traditional computer vision workflows. The company's founding philosophy was straightforward: video requires its own native architecture because meaning emerges through motion, sound, sequence, and context.

That philosophy shaped both the company and its name. TwelveLabs originated from an initial research team of 12 specialists spanning language models, machine learning, perception, and video research. The result is a company positioned less as a feature vendor and more as infrastructure for organizations seeking to operationalize video at scale.

Why TwelveLabs Matters Right Now

Enterprise AI conversations have largely focused on text, yet many organizations sit on enormous libraries of video that become difficult to use after they are created. Sports leagues archive decades of footage, media companies maintain extensive production libraries, governments store surveillance and evidence repositories, and enterprises record meetings, training sessions, product demonstrations, and operational processes every day. Much of that becomes dark data because locating a specific moment within thousands of hours of video remains challenging.

TwelveLabs addresses that problem by making video searchable and understandable through models built specifically for video. Its platform is designed to interpret what appears on screen, what is being said, what changes over time, and how events unfold across an entire sequence. For organizations managing large video archives, that transforms video from passive storage into searchable operational knowledge.

The Technology Behind the Platform

TwelveLabs develops proprietary multimodal foundation models for enterprise video understanding. Its model family includes Marengo for embedding and search, Pegasus for video understanding and generation, and Jockey as a research-preview model for reasoning across video workflows. AWS describes Marengo and Pegasus in Amazon Bedrock as managed models that enable organizations to build generative AI applications around video without managing the underlying infrastructure themselves.

That distribution strategy matters for enterprise adoption. TwelveLabs' Amazon Bedrock integration documentation directs developers toward Marengo for embeddings and Pegasus for video analysis, while its Bedrock blog explains how the integration supports search, analysis, and metadata generation across enterprise video pipelines. In practical terms, TwelveLabs is working to make video intelligence deployable within existing enterprise AI workflows rather than limiting it to specialized research environments.

Market Context

Video has become the dominant digital medium across nearly every industry, but the software infrastructure surrounding it continues to lag. Media organizations, sports broadcasters, public-sector agencies, advertisers, security teams, and enterprise operators all generate video containing valuable operational insight. The challenge is that most existing systems can reliably store video but cannot reliably understand it.

As multimodal AI continues to mature, buyers are no longer asking only whether AI can summarize a clip. They increasingly want to know how quickly video understanding can be integrated into production workflows while meeting enterprise requirements for governance, scale, and security. That shift creates an opportunity for infrastructure providers focused on video-native intelligence instead of treating video as another file attached to a text-based system.

Leadership and Strategic Positioning

Jae Lee serves as co-founder and CEO, alongside co-founders Aiden Lee, Dave Chung, and Soyoung Lee. Together, the founding team gives TwelveLabs a cross-border technical foundation spanning San Francisco and Seoul, combining expertise in applied AI, video understanding, and enterprise software. That blend reflects the market the company is pursuing: frontier AI research capable of supporting real-world enterprise workflows.

The company's investor network provides another meaningful signal. Publicly available information and investor profiles connect TwelveLabs with AI, media, infrastructure, and venture leaders, while Index Ventures includes the company in its portfolio. The broader pattern is clear: experienced investors and technical operators increasingly view video intelligence as foundational infrastructure rather than a specialized media tool.

Funding and Industry Validation

Bloomberg reported that TwelveLabs raised $100M from investors including Amazon-related funds, NEA, and NAVER Ventures. The financing reinforces the company's position in a market where video search, semantic analysis, and multimodal AI are becoming strategic priorities for large organizations. While detailed customer metrics and financial performance have not been publicly disclosed through primary company sources, the funding itself represents a meaningful validation signal for an enterprise AI infrastructure company.

The World Economic Forum has also profiled TwelveLabs for its work helping organizations unlock value from video archives through semantic search, summarization, analysis, and question answering. That recognition places the company within a broader enterprise AI adoption story. TwelveLabs is not simply making video easier to search. It is working to make video a usable organizational knowledge layer.

Why Hiring Momentum Matters

TwelveLabs continues hiring across technical and operating functions, with its official careers page listing opportunities in AI engineering, research, backend engineering, product, and related disciplines. For market observers, hiring activity is more than a recruiting update. It indicates continued investment in the research, engineering, and execution required to transform multimodal models into enterprise-grade products.

That matters because AI infrastructure companies do not succeed on demonstrations alone. They must simultaneously advance model quality, developer experience, enterprise trust, integrations, and commercial execution. TwelveLabs' hiring activity reinforces the same thesis reflected in its product strategy: enterprise video intelligence is becoming a durable market rather than a narrowly defined technical category.

The Bigger Industry Shift

Artificial intelligence spent its first major wave organizing language. The next wave is organizing perception, where images, audio, and video become searchable, analyzable, and usable inside enterprise systems. TwelveLabs is building specifically for that transition by focusing on video-native intelligence rather than treating video as a secondary input to text-first models.

Whether the application involves media production, sports analytics, enterprise knowledge management, public-sector operations, advertising, or security, the underlying challenge remains the same. Organizations need to transform massive video libraries into actionable intelligence without forcing people to manually review countless hours of footage. That is the market TwelveLabs is pursuing, and it appears well positioned as enterprise AI expands beyond text into full multimodal understanding.