Vibe-Coded to Production-Ready: Closing the Last Mile with Catalyst and Apptics - Vignesh Krishnan

In this session, Vignesh discusses the advancements in AI, particularly in application development. He shares an example of an app built using AI tools like Claude and the Catalyst platform. Vignesh explains that while AI significantly reduces the time taken to create applications, it does not address all the critical questions surrounding deployment and production readiness. He emphasizes the importance of Zoho's Catalyst, a platform as a service that offers comprehensive tools for building, deploying, and managing applications. Catalyst handles infrastructure, scaling, and other backend tasks, allowing developers to focus on functionality. Vignesh also introduces Apptics, an analytics platform that provides insights into user interactions and app performance, essential for understanding how applications are used in real-world scenarios.

Hi everyone. I'm Vignesh and I've been with Zoho for around 10 years now. I started as a software engineer and now I basically take care of DevRel for a platform called as Catalyst. So, AI, right? Everyone spoke about AI in the morning. I'll also talk about AI and probably everyone else, all the other speakers are going to talk about AI for the next couple of days here

as well. And all that hype and interest is well deserved because despite some skepticism and questions, if there is one thing that AI has really proven to be good at is in generating code and building the applications. So, um as we all know, with AI the time it is now taking to build an app has reduced drastically. So, something that probably took you like weeks or days,

you can probably build it in a few minutes today with how advanced these AI models have been building. So, let's take a little example. So, let's assume I work at a company that organizes events similar to this one and my boss comes and tells me, "Hey Vignesh, I want you to build a little app where people scan a QR code and the app displays a list of

all the talks going on today and they basically pick the talk that they attended, give a rating to it and then submit their feedback if any." So, building this is really easy, right? You basically go ahead and open cursor or your cloud code or anything and you give it a prompt and that's it, you have an app. And that's exactly what I did, too. So, what I

did was I gave a prompt to Claude, and I basically used the Claude VS Code extension. And then it gave me a nice little app that looks something like this. Let me just start this up. And you basically get the session list and then a result section to see what people have voted on. And maybe a little about section of the app. Again, the prompt is nothing

special. You know, it's just a little prompt that I gave to the AI to build this app. And the AI chose I know Node.js and Express for the back end. And then for database, it chose like a local JSON file to store in the data. And then for the front end, it picked vanilla HTML, CSS, and JavaScript. So, this is great. But, you know, in 10 minutes

I basically had a working app where the AI generated the whole thing, including the form and then the live results and everything. But, this is cool only if this was a hackathon. But then this was not a hackathon. This was something my boss basically asked me to build. And in that scenario, it works on my machine is not an excuse that I can give. So, the AI

gave me the app, but it not gave me answers for some critical questions, like where does this app run? And what happens when 200 or 500 people hit the app at the same time? And how do I know if it crashes and why? And how do I know which features of the app people actually use? And how do I know if they drop off before submitting? So,

of course this was like a very simple use case, but imagine something more complex and something that requires more scale, like trying to process your e-commerce orders in bulk where it's very time critical and you cannot afford to have failures. So, in all of these scenarios, uh you know, these questions become even more important and something that you need to consider. while the time taken to build

an app has drastically reduced, but the questions and the decisions that you still have to answer before actually shipping an app continues to remain the same because these are the critical questions. So, the bottleneck has moved away from the building part to actually shipping it to a production-ready environment. Now, I'm sure you can all see where I'm getting at, but before that, I would like to maybe

give a little introduction to Zoho as well. How many of you here have heard Zoho? Do you guys have any idea of what we do? Okay, that's good to see. So, for those who don't know, Zoho is a company that makes business software, and we have over 55 plus apps covering from, you know, finance, CRM, accounting to legal tools. So, basically everything that covers the business application

spectrum. we have over 150 million users using our applications actively every day, and from from over 150 countries, and we have data centers in 16 locations around the globe. So, and and again, all of these apps are integrated and they talk to each other. So, to operate at the scale that we do, and to kind of handle the load that, you know, that comes along with the

scale, we had to, you know, invest heavily in our own tech stack. And by that, I mean everything that you see here right from the bare metal to the application layer is all built and owned by Zoho. So, when Zoho started the cloud players, so if I remember correctly, they were not as massive or at the scale that they are in now. So, our early engineering teams

all had to figure this out. They had to figure out authentication at scale, load balancing, and everything. So, right from the bare metal to the application layer, we built it all ourselves. tech stack and this battle-tested infrastructure that we built is what we are now extending, uh you know, to developers like you around the world in the form of tools like Catalyst and Aptics. So, Catalyst, we'll

probably start with that. It is Zoho's platform as a service offering. So, we provide all the tools and services that a developer would need in order to build, deploy, and host their applications. So, for compute, we have our own functions as a service and computer as a service offering. We have our own object storage solution. We have our own monitoring and testing tools. And for front end,

we have our own, you know, static as well as dynamic hosting platform. So, every single thing that you would need to run and deploy your app, Catalyst has all of those inside one comprehensive platform. what I did was I gave I took the same prompt and then pasted it inside an empty Catalyst project folder that, you know, I had initialized and asked Cloud to give it. And

we provide SDKs, right? Whether you're writing it in Node.js, Python, Java, or whatever other programming languages, those SDKs live inside your project folder and without me explicitly having to prompt anything to the model, it automatically read the SDK and deployed it on Catalyst. So, it used Catalyst data store for the managed DB side of things and then uh the uh front-end hosting is also done on Catalyst.

And for all the compute, the AI moved it automatically to Catalyst serverless functions. So, maybe I'll give you a quick little demo of the uh Catalyst console as as well as the uh code base. So, this is how a Catalyst project looks like and you basically have uh separate folders for your client code and then your functions. And then uh the node modules is basically where your

uh SDK and all your other package dependencies reside. And the AI basically wrote a serverless function that handles all the uh API responses and then the logging of data in the databases. So, in the actual Catalyst console itself, this is how your project is going to look like and you have different Catalyst services. So, you can see that uh I have my function here, my advanced IO

function. And then for the back-end, I've used the cloud scale where I basically maintain a table of all the session ratings that people have provided. And one cool thing about Catalyst is uh we have something called the scopes and permissions that let you uh you know, manage the scopes and permissions that different users in your application have. So, we've made it very easy for you to build

a multi-tenant SaaS applications uh right from the Catalyst console itself with very minimal configuration. And then here we have the other components of Catalyst like we have uh no SQL for unstructured data and we have components for your authentication And uh if you want to do job scheduling like run job pools and crons, we have our job scheduling component. And we also have our own event bus

service as well. So if you want to listen for events that are happening in your app or other third-party applications, you can use Catalyst to listen to them and basically run you know a function that you want on top of it. So and then of course we have something called as Slate. Slate is basically our front-end deployment and hosting service. You connect your front-end repository whether it

is built in you know Next JS, React, Vue or whatever framework that it is that you use and you can have it set up in such a way that every time you merge something in GitHub Catalyst Slate will basically run the CI/CD pipeline for you and then deploy it and give you the URL. So in fact this presentation is something that I built and deployed in Slate

and I basically have a URL for this particular thing. So it's just quite easy to set up your front-end deployment with Slate as well. All right. So let's do something. I would like all of you to go ahead and scan this QR code pick a session and give it a rating that you want. But for my session, please give five stars. But yeah, please go ahead and

do it. So are you all able to access the application? So basically you'll have an app that is deployed and running in Catalyst basically. So, let me just open it up here. I have my deployment here. I'm just going to copy this URL paste it. And you can see the results. And you can see that it is being populated live right here. So, this is all running

on Catalyst. And basically, once you deploy your app, Catalyst takes care of all the scaling and infrastructure management and provisioning for you. So, that your focus is only on the app that you built while we handle the scaling and everything else. So, now going back to the Catalyst console, if you go to the Catalyst data store, this is the table, like I said, where we are storing

all the ratings. And you have the view here where you can see all the entries that uh uh you know, you all have given. Now, uh of course, this is a very simple proof-of-concept that I built to drive the idea uh home. If you want to add an authentication layer to it, you can go ahead and set up Catalyst authentication. And uh you can basically scale this

up with all these services that we provide. Yeah. Yeah, sure. Sure. So, uh maybe I'll get to the pricing part in the end. But to address your question, one thing that uh we put a lot of effort on is in developer experience. And like I said, all these components and services that we built, it was something that Zoho had to build and do the R&D for for

our own internal use first. So, by using this you are getting the same performance and capabilities of the, you know, very same battle tested infra that Zoho is using and providing for their customers. So, our key highlight would be our developer experience and also our experience in building, you know, B2B SaaS applications of massive scale for over 25 years. And when it comes to pricing, Catalyst has

two models. You can go for a subscription-based model, or you can also go for pay-as-you-go model. So, depending on whichever one you want, you can go ahead and use it. And because we are not using any other public cloud, and because this is not some wrapper on any of the other players in the market, we are able to pass on the cost benefits to our customers quite

heavily. So, if you want more of a specifics, maybe I can explain the pricing to you in detail after that. since you specifically asked about I can guarantee you that you will see big cost benefits. Yes, correct. And in Vercel, I believe they don't let you deploy commercial projects in their hobby plan. We don't have any such restrictions. And basically, it's your app. As long as you

sit within the free tier, you won't be charged for it. And one key thing is Catalyst does not charge for bandwidth at all. So, we only charge for the compute, basically. Since you specifically asked about Vercel. >> [clears throat] >> Yeah. One second. Yeah, I'm sorry. Come again? You have an order quantity quantity. Yeah, so do you mean for building apps for Catalyst? Not specifically. So, since

you know, you can use Codex Cloud Code or whatever it is that you want. And Catalyst will work out of the box with it because these models are now advanced enough that they can go through the SDK and understand what methods they have to use to make use of the platform features. So, the good thing here is that whatever model you works as long as it is

capable, Catalyst will work alongside that for you. And regarding the models part, Catalyst also has AI and ML capabilities built in. So, if you want to build your own predictive intelligence models either for recommendation or if you want to do turn prediction or something like that, we have a service that lets you build them as long as you bring in your So, the key thing about Quick

ML is that you don't have to spend a lot of time in the data pre-processing and cleaning up step because we provide a very easy-to-use drag-and-drop canvas interface of sorts. So, your focus is on the model and then in training it instead of spending a lot of time in the pre-processing stage. And on the LLM side, we also provide a few LLMs that you can use out

of the box. Say, if you don't want to use you know, say Claude or OpenAI's GPT models for your specific use cases, you can use some of these open-source models that we host and provide for your usage. So, our team is working on adding more models like GLM and others as well. So, you'll basically be hitting the Catalyst endpoint and you know, run your prompt to the

model and then you do whatever processing you want So, that probably depends on a case-by-case basis. But for front end, it's pretty easy. You just point your GitHub repository here instead of there. And then once you do the domain mapping, you're good to go pretty much. if you're using emergent or something, you're probably having some code that is specific to their platform. You'll probably have to make

changes to that. But if you have a project or a web service that is not tied to any platform and you just have a Docker image for it, for example. So, if you want to deploy that, we have App Sail, where you can put your Docker image on top of it. And we provide basically the compute for you. So, depending on how your existing app is, moving

to Catalyst will basically change. So, that's more of a case-by-case scenario. Yeah. Real time? So, like your project building and other things, yes, we do. I have a question. Uh one second. That gentleman asked a question. I'll just address that. Just So, you basically get an overview of the And then the breakdown of which component and which projects have consumed the most. And then you get further

reporting and all these analytics. So, yeah. Yeah, next question. Yeah, so when we leverage this product, will we be leveraging it on the whole from application layer to database to network? Or we can we can choose those capabilities that we would want? Because we would have already be using, you know, for example, Akamai for the network. Yeah. And we do not want to wish to move Uh

so, is there a way we can customize? >> Yes. So, you know, usually when customers come to us, we tell them Uh just pick little parts that you want to move first and see how Catalyst works for you. And then eventually they all, you know, migrate things piece by piece. So, maybe you don't want to move your current setup away from what you have because it works

great. >> Fine. But if there is a little microservice that you want to probably start or like an internal app or something that you want to build, I would highly recommend that you start there. And once your team is also more familiar with the platform, they'll also gain the confidence to tackle on the much bigger migration projects that come in later. Okay. And one one last So,

you said that it is definitely cost-benefiting. >> Yeah. And currently most of the enterprises they're going ahead with services provided by different cloud, you know, right? Uh has that analysis been made between those providers and your product that it is definitely beneficial? Yes. Yes. Okay. Uh so, we have that data. Maybe I can probably get it from the sales folks and then share it with you. Thank

you. Yeah. Sure. Thank you. So, like I said, this is how the Catalyst platform works and we were able to see the data coming in. And yeah. All right. So, the app is running great on Catalyst. But then that is not alone enough, right? You then need more insights on how it is working, whether users are using it the way you intended it to, what's the analytics

on that. So, these are also important questions. And for that I would also like to introduce another analytics application application analytics platform called as Apptics from Zoho where you basically get insights about your current sessions, the active users, and the different events that happen in the app. And then for marketers, they get valuable insights like funnels like how many people access the app, but how many actually

failed to give in you know the ratings in this example. And then you get further information like device and browser breakdown. Uh I will show you the the same project that we deployed now. I'll open the Apptics dashboard for let's refresh it. information. Okay, let me just change this to 14 days. Yeah, got it. So, we can see that 56 uh new devices were you know here

who accessed this application for the first time. And then you get insights on the stickiness of the app. And if you scroll further down, you get insights on what kind of devices they were from whether it was from a browser or if you integrated the Apptics SDK, which is very easy to use and integrate inside your mobile application, you get that kind of analytics here as well.

And here you can see some events that we have configured such as uh app the app opening of the app and then the session selected and then the navigation and all these events happening. So, if you're a marketer in the team and you want to know say how many people access the app and how many dropped off before actually adding a comment, you get that kind of

insights also here with Apptics. So, think of it like this. You you now get an idea of not just uh where the app is running and whether it is healthy and all those technical parts through Catalyst, you also get how your app is being used by your users also. And in this example, I have integrated Aptics web SDK inside the application, but Aptics is also full-fledged and

it has support for all the mobile platforms as well. So, if you are someone who is building iOS or Android apps, Catalyst has sorry Aptics has even more integration and it can provide you insights on why your app crashed, and you can also set up in-app feedback inside your application to collect feedback from your users and then further things like in-app ratings and whatnot. So, let me

just go back. So, we basically saw two ways of looking at your app. So, you can use Catalyst for figuring out how to run and deploy your app, whether it is serverless functions, managed DBs, or front-end hosting and deployment or auto scaling. And then you can use Aptics to get insights on how your app is actually being used by your customers either through events, funnels, user journeys,

or product insights. So, you know, AI today can get you 80% of the way and the last mile is the production part, right? Where exactly do you ship it? And the last 20% does not really have to take 80% of your time. And that is what we are here to help you with and we really hope you try out Catalyst as well as Aptics. So, feel free

to scan the QR code to get further information about both of these products. And if you have any questions, I'm here to take them. I also have my colleague as well from the Aptics team. And we have a booth, the Zoho booth. I highly recommend that you come say hi to us and we'd be happy to give you an in-depth demo. So, thanks everyone. >> [music]