Product builders around the globe continuously trade-off between three methods
When do you have to (re)-build your fashions and put money into iteration to get nice high quality? When do you have to borrow high-performance APIs and free your builders to make the subsequent large factor? And when do you have to purchase options to your drawback?
This week, we talked to VideoKen co-founder & CEO Vishnu Raned and his crew, to know how they traded-off between “construct” and “borrow.”
What’s VideoKen?
VideoKen is an AI-powered video interactivity resolution that transforms studying movies into interactive and immersive experiences. On this age, consideration is at a premium, and learners have entry to extra info than they’ll devour. “VideoKen’s function is to take learner engagement to the subsequent stage by making it simple for individuals to find and work together with movies,” says Vishnu. VideoKen affords automated video chaptering, in-video quizzes, deep video search, video analytics, closed captioning, video internet hosting and different options to make this interplay wealthy and impactful.
What occurs behind the scenes at VideoKen?
VideoKen’s resolution makes use of AI to trace, predict and enhance learner engagement utilizing in-video interactivity and actionable insights. The aim is to reduce the issue of poor engagement amongst learners.
Over 100+ studying leaders leverage VideoKen to assist learners interact and work together with movies utilizing an auto-generated navigable storyline, deep video search and in-video assessments. These leaders additionally faucet into VideoKen’s in-depth analytics to achieve insights into their learners’ preferences and habits, which is used to optimize and prioritize their studying content material.
“VideoKen’s clients have been profitable by driving 3x extra engagement and retention on their studying movies,” based on Vishnu.
VideoKen’s growth journey
VideoKen’s journey began by creating in-house know-how (now patented) for bettering engagement. On high of their 5 patents, VideoKen wanted to have the ability to supply a complete resolution to satisfy rising buyer wants. “The engineers had been keen to construct in-house fashions, but it surely turned out we couldn’t do all of it by ourselves,” says Vishnu.
Furthermore, VideoKen wanted an enough cloud computing infrastructure which led to exploring and experimenting with numerous choices earlier than deciding on Azure.
Growing in-house machine studying fashions for speech-to-text transcription
Jeril Sebastian, Head of Engineering at VideoKen, shared insights on utilizing a number of the utilized AI companies. “Most of our companies depend on VideoKen’s personal AI know-how. One factor that we do depend on externally, is Azure’s Cognitive Companies for transcribing movies.”
Jeril defined that earlier than leveraging Microsoft Azure and its Cognitive Companies, VideoKen was utilizing a mix of its proprietary speech engine together with some companies of one other cloud supplier. “It was taking on a whole lot of time and wasn’t possible for us to pursue this strategy long-term,” Jeril mentioned, including that the corporate was spending twice as a lot cash on the in-house mannequin in comparison with Azure’s speech API. “Despite the fact that every little thing was working nicely with our AI speech-to-text resolution, we would have liked to give attention to our key strengths and have our builders work on options which can be necessary to our clients.”
After discovering that internet hosting GPU cases and coaching fashions in-house was taking on too many assets, VideoKen determined to check totally different choices, with Azure’s speech to textual content proving an ideal match. “Speech to textual content is a single constructing block of a a lot greater infrastructure, so we couldn’t afford to lose a whole lot of time and vitality, regardless of being glad with the general efficiency of our home-grown know-how. Subsequently, changing it with Azure was probably the most logical step for us.”
VideoKen’s journey post-Utilized AI
Video transcription has turn into a norm these days. Some of the necessary features of constructing movies extra immersive is making certain that learners don’t miss any necessary info. Having realized Azure Cognitive Companies’ energy, Jeril and the engineering crew managed to chop the price of video transcription by 50%.
“We used to have an ML engineer simply to coach the mannequin, enhance it, run the GPU cases, and so forth,” Jeril mentioned. “This was a full-time job for one particular person, and now it takes just some hours per week for me to handle that total pipeline. Our ML engineer now has extra time to give attention to researching and upgrading different companies provided by VideoKen, as a substitute of doing the grunt work to maintain our in-house transcription mannequin aggressive.”
Processing 1000’s of movies with Azure
Aside from Cognitive Companies, VideoKen additionally depends on different Azure advantages. “We use two particular items of know-how. One known as Azure Capabilities, and the opposite is Azure Container Cases,” defined Jeril. He added that the 2 tech items allowed VideoKen to scale up and down elastically, “We get 1000’s of movies at a time from our clients, so fast scalability is crucial.”
By counting on Azure Container Cases and Azure Capabilities, VideoKen has managed to arrange a pipeline which allowed the corporate to scale. “When there aren’t any movies to course of (theoretically), it prices us nothing,” Jeril mentioned. “However, when there are literally thousands of movies coming in, VideoKen scales up robotically and is ready to deal with that load.” At full capability, VideoKen can course of roughly a thousand hours of video per day.
Jeril shared his pleasure with Capabilities and Container Cases. “I actually get pleasure from working with Capabilities, because it has saved us a whole lot of time. This can be a realm of know-how referred to as serverless, and it has made administration and upkeep very easy.” Furthermore, Jeril acknowledged that Container Cases helped VideoKen’s scalability in a singular method, on condition that such know-how doesn’t exist on different cloud companies.
The long run seems to be brilliant for VideoKen’s partnership with Microsoft, “Proper now, we’re proud of Azure, and the plan is to stick with the service for a very long time,” mentioned Jeril.
He expressed his satisfaction with Azure’s transparency relating to prices. “They’re attempting their finest to indicate precisely how and the place our cash is spent for Azure’s companies. At any time when I go to the fee administration part, I can see precisely the place we’re spending cash and the way a lot is being spent.” The principle benefit of this, based on VideoKen’s Head of Engineering, is that the corporate could make higher selections on the right way to use the Azure funds.
To get began with Microsoft for Startups Founders Hub, join right here.
