Inside the VALT: Treating Video as a First-Class Data Source with Wowza's Barry Owen

In this episode of On Record, hosts Mike Anzalone and Carmen Haak speak with Barry Owen, Chief Solutions Architect at Wowza, the streaming engine technology that powers the video component of the VALT platform.  Barry took an engineering path from the start, working as a programmer on image processing products before video work drew his interest. He knew Wowza's founders from his previous work at Adobe, and the timing lined up for him to join the company, where he began in engineering. He has since moved into his current role as Chief Solutions Architect, an outward-facing position focused on understanding customer needs and determining how video can address them.  The episode opens with Barry's take on where the streaming industry stands today. Getting live video out at scale, securely and reliably, is now a largely solved problem, and the newer challenge is drawing real-time insight out of video that would otherwise sit unwatched in storage. He points to Wowza's recent AI product launch and the beta program that preceded it, including a use case from an offshore monitoring customer that inspects oil rigs and other harsh-environment infrastructure. AI models trained on remote camera feeds can flag events such as bubbles escaping from a pipe or a pressure gauge reading that is too high, generating a real-time alert instead of requiring a person to watch video around the clock. Barry notes that the same approach generalizes to clinical simulation settings, where a model could flag whether a patient is out of an expected position at the wrong time.  On what Wowza Streaming Engine actually does, Barry describes an end-to-end platform that takes in video from sources like cameras, files, or satellite links, processes and packages it for delivery to any device in near real time, and layers in security such as encryption and access control. It is built to be API-driven and is typically embedded inside other products, which is how it functions within VALT rather than being sold as a standalone tool. On scalability, Barry explains that Streaming Engine is designed to take advantage of underlying hardware, including GPUs, to process high camera counts efficiently, and that it can deploy on premises, on edge devices, in the cloud, or in some combination. The right architecture depends on the use case: a customer whose stream count fluctuates sharply, such as jumping from 100 to 1,000 streams on weekends, needs a different setup than one running a small number of streams out to a large audience.  Among the more distinctive deployments Barry has worked on: streaming live video and telemetry from Formula One cars to the pits, supporting NASA's Artemis program with live video from the moon, and enabling E911 operators to receive real-time video from callers. He frames the broader shift underway as treating video as a first-class data source rather than a passive recording, and connects this to the healthcare simulation market, where video functions as evidence supporting learner competency. Wowza's platform is designed to be open, so organizations can add or train models suited to their own use case, such as a clinical model that expects one event to follow another within a set time window and flags it when that does not happen. On working in sensitive environments such as healthcare, law enforcement, and research, Barry emphasizes that video analysis with Streaming Engine and Wowza's Video Intelligence framework runs on premises or at the edge rather than being sent to the cloud, which allows deployment inside existing hospital networks, police stations, and other secure environments without giving up control of the video.  Asked where Wowza's AI capabilities are headed over the next several years, Barry traces a path from basic object detection, such as counting vehicles, toward more predictive capabilities like anticipating a traffic jam or accident from a scene, and toward synthetic video detection built with a model from Nvidia that flags AI-generated video so organizations such as news outlets can handle it differently than user-generated or social content. He also discusses context detection, which can be used to keep advertising or content transitions from clashing tonally with surrounding footage, and to support content moderation at scale. Asked what sets Wowza Streaming Engine apart, Barry points to its API-first design, dating back to the product's original launch in 2007, and its extensibility. As an example, he describes using Claude to help him write code again after a decade away from it, which let him add a new protocol called MoQ, short for Media over QUIC, to Streaming Engine using only its public APIs, without modifying the core product. He explains that MoQ is built on QUIC, the UDP-based transport behind HTTP/3, and is designed to address the latency issues inherent to TCP-based protocols like WebRTC, while also supporting features WebRTC was not built for, including accessibility and DRM.  Toward the end of the conversation, Barry offers practical advice for organizations upgrading or starting a video streaming initiative. His recommendation is to pick one clear use case, get it working end to end, and only then think about scale, using existing cameras and servers wherever possible rather than assuming new infrastructure is required. He notes that Wowza's deployment options range from small edge devices to multi-server GPU setups, and that most streams today are ultimately viewed on mobile devices, which should inform how output is optimized. His closing thought for listeners already streaming is to look at how that video is actually being used, and to consider whether new AI and video intelligence tools could help extract more value from existing infrastructure and past investment.  Barry Owen can be reached on LinkedIn or through his posts on the Wowza blog.