The financial sector has special requirements when it comes to communications. This has created a small community of companies that service this sector at high cost and with traditional products. When vendor replacement comes at high cost there is always the risk of products staying behind the technology curve.
WebRTC removes barriers to entry in many sectors, apparently this is the case also for the financial sector. In a previous interview I had, Leo Papadopoulos from Cloud9 Technologies described their service and market disruption. In this interview I want to revisit the financial sector, this time having Joseph Heenan, VP Engineering of GreenKey Technologies, describing their service and how GreenKey is moving the cheese for the incumbents.
After speaking with Joe it looks as if the incumbent communication providers for the financial sector have a lot to worry about as new innovative, cloud based and affordable services are replacing their cash cow products.
Q. What is GreenKey Tech all about?
Green Key Technologies was founded in 2014. Our initial product was back end infrastructure for brokers and traders that need push-to-talk capabilities to negotiate trades in fast-moving markets. In 2015, our front end was built, initially with Java, then leveraging HTML5. Since then we have continued to evolve.
In simple terms, every phone in the financial industry should be powered by GreenKey. The financial markets run on voice (price discovery, market color and insight, any large or complicated order, even regular trades). At GreenKey, we bring the art and science of voice together. We take the efficiency and nuance of voice and pair it with the power of digital functionality and analytics. And we make voice compliant.
GreenKey is the voice-driven collaboration platform for financial markets participants: enhancing compliance and enabling sophisticated data analytics.
Q. What is special about GreenKey’s offering that disrupts the financial market’s collaboration industry?
At GreenKey, we are enabling a paradigm shift in the economics and governance of trader voice communication. We offer a solution to replace the traditional high cost ($10,000+ per unit!) and low flexibility hardware “turret” phone systems with a secure, WebRTC-enabled application that comes with a host of value added services. We are 1/20th the cost of incumbent solutions with none of the headache.
Q. Tell us about the architecture of your service and the technologies being used
We use FreeSwitch heavily as our back-end media server for audio multiplexing and transcoding; we also have an AngularJS/HTML5 frontend. FreeSwitch is scaled horizontally to handle additional load via our management application. We use Docker and Ansible heavily for deployments so the app can either be run on premise if desired or on one (or more) clouds in our hosted environment. Writing our front-end as an HTML5 app has proven to be a key strategic decision as it’s allowed us to incorporate features at a much faster pace across multiple platforms – our app is usable today on Windows, Mac, Linux, Android and iOS – with a Protractor-based automated test suite that helps us ensure it remains reliable and performant with each release.
Q. How are you using WebRTC? Is it in the browser or are you compiling the open source?
For the desktop we use an Electron and Chromium-based container provided by our partner www.openfin.co. For iOS we do compile the native WebRTC libraries from the webrtc.org source repository. We love the high-definition audio and echo cancellation we get “out of box” from the WebRTC codebase.
Q. What are you using for signaling and why?
We primarily use SIP over WebSockets for core telephony signaling (although parts of our application rely increasingly on newer protocols like Verto). We feel that SIP is the de-facto standard for interoperability with existing turret hardware, which makes it a logical choice for our core signaling protocol.
Q. How did you handle WebRTC on mobile?
On Android we are able to very simply wrap our app in a WebView. On iOS, it’s a bit more difficult, but not much – we can still use Cordova and our responsive AngularJS-based UI, but using a modified PhoneGap plugin that leverages the WebRTC project’s iOS reference libraries to handle audio.
Q. What are the challenges to implement push to talk using WebRTC?
Push to talk requires instant connectivity, which means a highly optimized backend and a strong ability to scale horizontally.
Q. What would you want to be different or added to WebRTC?
There isn’t a good reference implementation (that I know of) for implementing a WebRTC-based real-time streaming audio service at scale (1 presenter, many listeners).
I’m also hoping Green Key can engage with the relevant working groups to come up with a tighter integration between the WebRTC spec and the WebSpeech API. We’ve spent a lot of time engineering a real-time transcription engine for WebRTC-sourced audio that (for the financial markets domain) is both faster than Google’s for online decoding and more accurate. Accuracy related to domain specific data is often more than 97% whereas same data in Google is around 87-95% – our financial markets knowledge and ability to define custom grammars for recognition have been key advantages here.
Q. What’s next for GreenKey?
We just moved into a bigger headquarters office in Chicago and continue to expand – we are deployed at over 350 firms now and in POC at many of the world’s largest banks and brokerages. We’re working hard on our mission of delivering significant cost savings over legacy hardware to our customers, and improving customer productivity and compliance by turning voice into actionable text data in real-time.