Built for Scale: Pypes

What are Pypes?

Julie: Pypes are a two-way messaging carrier that interface between the end-user consumer and the conversational interface with the engagement automation subsystem, and it allows for a secure and rich messaging solution between the two.

Rahul: So, Pype is a tenant for a customer, right, it allows a private set of customer data to live within our multi-tenant system.

Why are Pypes important?

Julie: Pypes are important because it allows a private and secure exchange of information between the end-user consumer and the backend system, so they’re able to perform any transaction that they’re looking to achieve.

Rahul: Pypes are important because it helps segregate data between multiple customers and each Pype is protected using Amazon keys and what that does is prevents customers from ever commingling in data

Aaron: Security is incredibly important – for a few reasons, so, first we do a lot of work with companies like insurance companies, for example, and they have very secure customer data they don’t want spilled. They do not want individuals who don’t have access to that data to be able to obtain it, so the Pype allows a secure method to authenticate a user. Without a Pype you’re not really going to have a secure method for obtaining specific customer information and so they’re really critical to ensuring that security and ensuring that we have the right information for the right customers.

Why was the platform built in Erlang?

Rahul:  It allows for our system to scale up and down and it can handle a lot of messages at the same time. The biggest load on Erlang is normally when a message starts after that it can guide a message, it tracks messages, and it makes sure messages are completely received at each destination.

Nick: The reason we chose Erlang is because it’s really extremely well-suited for building both scalable and fault-tolerant systems. Erlang comes with all sorts of features right out of the box that are extremely good for both scalability and reliability. And those are two really important things to us, we’re cloud hosted, we want to be able to scale up to many customers and never have to worry about problems from one customer affecting another. So, besides just being reliable Erlang is also really good for fault isolation, but isolation is about acknowledging that even if a problem does happen you need to make sure it doesn’t affect any other customers or any other chats and that’s really important to us and so that’s why we chose Erlang.

How are Pypes secure?

Ray: Pypes are secured in multiple ways. The first as, Pypestream as a SAS provider, we offer our service in the cloud and we use Amazon Web Services for our infrastructure, so first of all, we rely on Amazon to provide all of the physical data center security that our solution is run in, since everything is hosted in their environments. The second part is, in terms of the data that traverses through our solution, it is secured using encryption. So, we use encryption of both data in motion, or data in transit, as well as encryption for data at rest. We also support redaction, so that if anything is sensitive that is entered in the process of a chat, for example a social security number or some medical record information, that data is redacted so that doesn’t appear in any of the records that Pypestream generates of the chat.

How does encryption work?

Rahul: So, encryption uses different keys at different levels, a user may have a public key and that a company may have a private key, the public key is used to send the data and a private key is used to un-encrypt that data, only one person normally has a private key.

Ray: So, the purpose of encryption is to make sure that sensitive data is not legible by an attacker. So, if somebody gets illicit access to that data they can’t use it, they can’t read it, they can’t benefit from it, essentially makes the data useless. Um, so, there are multiple types of encryption there is encryption for data that’s in transit, or data that’s moving across the Internet, there’s also encryption for data at rest. Both rely on very sophisticated math to make sure that only the people who have access to the keys can actually view that data. And the keys are very, very, very long numbers that are you know make it very difficult for somebody to be able to decrypt data that they don’t directly have the keys for. So the whole purpose is to make it essentially, make that date of valueless

What compliance standards do Pypes adhere to?

Rahul: Pypestream adheres to PCI and HIPAA in the US, as well as GDPR in Europe.

Ray: And the important thing about those compliances are that they ensure that we have the right processes in place from a security perspective to ensure that our data is properly protected and that the right controls are in place from an employee perspective, from a customer perspective, that access is never available illicitly to any sensitive data.

How are Pypes different than other ways companies deploy messaging solutions?

Rahul: Pypes allow us to have multiple solutions running on the same Pype, so they are truly a tenant per customer. So, a customer may have five solutions running on the same Pype, what that allows them to do is reuse those agents on the same Pype but basically get the efficiency of their company all using the same Pype but be able to do multiple different solutions.

Julie: So it allows for information to securely be exchanged and confidential information can be redacted both when it’s in transit and also at rest. So, information is very secure via the Pypes.

What are the benefits of Pypestream being cloud-hosted?

Rahul: Pypes that are hosted in the cloud allow for a company to not have to worry about development or building of the infrastructure, as well as they get economies of scale by sharing the infrastructure with other customers.

Ray:  Pypestream’s responsibility is to, you know, provide software updates – they all happen in the background – to add new features, so it’s something that, it’s very easy for a customer to consume and so from an industry perspective we’re seeing, you know, a movement towards software as a service in general. They can start offering a conversational AI service without procuring equipment, without having to train staff on the solution and development of the service, without having to apply updates, without applying security updates, it’s all managed by Pypestream, we offer a continuous development environment where we’re continually adding new features and incorporating them in the background. So it just makes it a very simple and easy service for customers to consume.

Why do enterprises prefer deploying conversational AI via Pypes?

Rahul: Because Pypes are a lot more flexible, right, Pypes allow for encryption, they allow to redact data, they give customers the ability to drive business solutions via Pypes where, if they were to do it in another, another, if they were to do it using another vendor, right, they have to build some of the other pieces that, that aren’t there, right?

Aaron:  So, we provide your customers with a very nice-looking, easy to use, intuitive interface, that they can address all kinds of needs with.

Check out the other videos in the Built for Scale series: Dovetail and Conversational Interface