← The Digital Executive
The Digital Executive
Pratik Balar on Building Trustless Infrastructure and Empowering Developers in Web3 | Ep 1088
00:14:26
Loading summary
Transcript13 lines
- [00:00]
A
Foreign welcome to Coruscant Technologies, home of the Digital Executive Podcast. Welcome to the Digital Executive. Today's guest is Pratik Bilar. Pratik Bilar is the co founder and tech lead at node ops, a web3 infrastructure company building decentralized compute systems that power the future of AI and cloud services. With over 3 years of hands on experience, Pratik has architected and managed hybrid enterprise grade infrastructure across Tier 1 cloud providers and bare metal environments, optimizing for scalability, uptime and performance at production scale. He's deeply invested in building infrastructure that lowers the barrier to adoption for developers while reinforcing the long term sustainability of decentralized protocols. Well, good afternoon Prateek, welcome to the show.
- [00:52]
B
Good evening.
- [00:54]
A
I appreciate that you're in Bangalore, India right now and so it's early morning for me but it's evening for you so I appreciate you making the time. Pratik, let's jump into your first question. You've been instrumental in building Node Ops decentralized compute systems. How do you envision Deepin 2.0 transforming the future of AI and cloud services?
- [01:14]
B
Thank you for question. I mean I think it's a good question. When it comes to Deepin, all I can think of is like scalability, cost efficiency and privacy. So deep into point oh, we'll transform AI and cloud services by creating, you know, more scalable, cost effective and ecosystem. Of course by leveraging blockchain Token incentives, Deepin 2.0 also enables anyone to contribute idle compute resources like GPUs and storage, reducing reliance on centralized providers like a tier one cloud provider. I won't name anyone but yeah, you got the idea. This democratized like access to AI infrastructure, slashing cost to buy 17 to 90% in some of the cases compared to traditional cloud providers, enhancing scalability to global permissionless network. All the deeping anybody can spin up a node from any region of the world and they can be a part of the whole DPIN cloud or something like that. Deepen 2.0 integrates real time data from IoT and edge devices as well as we all know like Geodnet and all are doing enabling AI models to train and infer on diverse high quality data set. So yeah, this is kind of how our and how this is transforming the future of AI and cloud services.
- [02:33]
A
That's amazing and I appreciate you breaking out Deepin. Obviously that's so important now in the decentralized world. What I took away from this is obviously the scalability and the fact that it's cost effective you know, leveraging blockchain technology for decentralized services. And you mentioned providing token incentives which I think is awesome. And Deepan 2.0 allows advanced integration for example, you mentioned IoT devices, cloud, et cetera. So I really appreciate you highlighting that for us. Pratik Node Ops focuses on building infrastructure that is open and trustless with what are the biggest technical and philosophical challenges in designing such systems at scale?
- [03:13]
B
Building a lot of smaller right to be honest, like open and trustless infrastructure at scale like a lot of faces technical challenges such as ensuring scalability, low latency consensus and robust security against threats like Sibyl and DDoS which we have faced during our testnet and we are continuously facing ddos on our mainnet as well. I think last month we got somewhere around 7 like 7 ish $30. Luckily we have a lot of things that we have to take care of. Interoperability across heterogeneous nodes and maintaining data integrity in decentralized environment are also crucial. We have to create a lot of in house tooling to secure node to node connectivity by leveraging something like technology which is being used on a lot of tech stack like mtls, which is like mutual tier stuff. A TLS certificate or certificate which will be given to both parties and both parties can communicate and authorize each other's communication using this single tls. So this TLS will be mutual among both parties or something. Things like that we can do to, you know, secure node to node connectivity. This is again one of the examples. Second would be managing state across the nodes. So let's say if I'm running Polygon, let's say I'm a node Ops orchestration and I'm running Polygon network or let's say Polygon validator onto your nodes and for some reason, for some XYZ reason your node is down or under maintenance state. So now I not only have to reschedule this node but also to take whole state which is like blockchain data to the other nodes as well. We have to thinking we have to keep copying or we have to keep the backup of the state and we have to keep snapshotting this state and then migrate these two some other nodes. So these kind of challenges again I would say creating everything in a non cloud or like in a generic fashion was the challenge for us especially apart from this we are also researching around privacy computation with and without T. With T. If I review with T that would be running something enclave environment where your node mean your node's private key. Let's say your private key will be under T so nobody, not even a machine owner can see that private key and your node will continuously do its thing. It could be like chucker node or it could be a validator node or it could be a ZK WiFi node, something like that. We are also the thing T is not that scalable so we are also researching something around without T where we do something like aphe like fully homophobic encryption where artists can or some nodes can directly do computation on top of encrypted data without revealing what's in that data. Something like that. So those kind of things we are doing these kind of technical challenges we have faced at scale and this is something we are researching now.
- [06:01]
A
Thank you and I really appreciate you helping us understand that a little bit. I know the vision was to build open and trustless as we like to do in the decentralized world. Having that secure node to node connectivity was important. I highlighted that. And then the managing that uptime between nodes is easier in your environment. I appreciate that. Pratik, you've worked across cloud giants and bare metal environments with what are the trade offs between centralized and decentralized infrastructure from a DevSecOps perspective.
- [06:29]
B
So that's a really interesting question which I get a lot during like meetups and all from a DevSecOps perspective which stands for developer security and operational centralized infrastructure like cloud giants Tier 1 clouds offer streamlined management, robust security tools, rapid scalability but introduce vendor locking, higher cost and single point of a year like few weeks ago when I guess GCP was facing some issue and like majority of region was down due to some tail on terror side pushes, some bad commits and that also took down a lot of other things like cloudflare and all which is a major threat on centralized worlds. I don't know like people have seen but that test node also was not down even though we are utilizing gcp. And also that's the beauty of I think deep thing. By utilizing the bare metal decentralized infrastructure we can avoid this totally. And as NodOps is doing we are utilizing we utilize multi cloud hybrid setup for everything which provides like a greater control cost efficiency, resilience against censorship but also faces challenges with consistent security and for client complex and slower deployment cycle. So it requires of course higher maintenance. But now you don't have a single place to maintain everything now it's catered and then you have to manage all those things. But centralized system simplify compliance and monitoring as well while decentralized setup requires a distributed trust mechanism and advanced consensus protocol as well in some cases which increase this operational complexity and balancing security, scalability and agility remains critical in choosing the optimal infrastructure for a specific use case. In our case it's a bare metal flux mips of other tier 1 cloud. I think those are the like trade offs between centralizing decentralized infrastructure from DevSecOps per se.
- [08:26]
A
Thank you, I appreciate that. You know had someone else recently on here that was a big proponent of decentralized cloud and obviously you talked about GCP last week. See this every week with the major cloud providers, they're more vulnerable to attacks, they may have a bad code push, then they've got to reverse that out and it affects everybody and it's hard. But I like how you're using when we're talking about DevSecOps, you're using the decentralized environment. You know you can streamline your management, there's more cost efficiency, it's secure and then of course that distributed trust environment you talked about. So I appreciate you highlighting that. Pratik, Last question of the day. Deepin 2.0 aims to be more developer friendly. What specific steps are you taking at Node Ops to reduce the friction for developers entering the Web3 space?
- [09:12]
B
Again, good question. So we have a few free testing NoDOPS cloud for devs and for beginner dev one of them would be. I want to highlight few of them. One of them would be our template based deployment where any beginner dev or any dev can create a YAML based template of their favorite code. So let's say if you want to spin up a uniswap tab and then you'll just create a YAML file of it and then you'll put it into NodOps Cloud. The YAML file is very simple, it's just a wrapper of docker compose or kubernetes manifest which is like the way how you want to deploy your stuff. That's what is defined in the YAML and then your deployment will be taken care by NoDOPS Cloud Orchestrator and then we'll schedule it, we'll maintain it, we'll show you uptime, we'll show you the traffic, we'll show you around how much resources are being used it's utilizing and all. And we'll give you a link which will be enforced by all these security tools like you won't face like any kind of ddos will take care of it. Your small depth won't go down even if you face any DDoS and all. So those all things are like Backed by all the orchestration things that we have created around it. And we are soon releasing our replica feature on top of it which allows you to deploy your favorite code onto some particular region layer region. So let's say if you are thinking that okay, you are getting more traffic from usa, we will give you the flexibility of creating multiple services into USA to cater that region of the world. So things like that we are also planning. Second would be AI Sandbox. AI Sandbox is really interesting application that we have developed or feature that we have developed onto cloud. It's like seeing the user of various code like Sandbox with all the necessary toolings like Node js, Rust and all the old fancy programming language being inbuilt again. Into that sandbox we have this GPU support so you'll have whatever Nvidia graphics driver, whatever you need to run your code. And this, this sandbox will have everything in order to start as a dev. So now you don't have to configure everything into your local, you have to download this packet, that packet and you have to set up this programming language that you just spin up AI Sandbox and everything will be set up. You just start developing at that point. And then you can even utilize VS code extensions like let's say GitHub, Copilot and all to do your bytecoding as well. So it will checks of that white coding checklist as well. On top of this you have some features like package manager. So let's say if your code is not compatible with latest Node JS version, you have to drill it down to something like let's say second tier, like let's say 22 or 20 version. So you can just do a single command and drill down your programming language version as well. We have something called Chrome Doc feature where you don't have to visit your AI sandbox every time, you just have to open it into Chrome and Chrome will show you pop up that okay, add it into your dock and if you are a Mac or Windows person just add it into your dock and by using single click you'll have this whole your ID programming ID into your local. We have also this feature of port tunnels where you can open any port from the remote machines to your local. So let's say if your remote machine running some kind of AI model, let's say 200 billion parameter AI model and you want to access it to from local, you can just do a port tunnel and nod Ops will create an end to end port from your local machine to the remote Instance where the whole thing is running 200 billion parameter model is running and then we'll just. You can do whatever you want to use that tunnel for. So things like this is then last would be RPCs in AI APIs. So NodOps has been continuously contributing towards RPCs where by creating public RPC so you'll be able to consume public RPCs of hyperliquid. And then the second would be movement labs and the self chain, things like that. And we also enables AI APIs. So let's say if you want to self host your grand model or let's say you want to self host your own Gemini 30 billion model will we do it just DM us? Or you can also create it using your template onto our Node O cloud. Things like this like widely available for any beginner dev or anyone who is entering into web3 space can directly consume it and create it will give the whole first five step and you'll have all the resources you want and then you can just kickstart your journey into Web3.
- [13:37]
A
Thank you, I really appreciate that. A lot of features there that you unpacked. I'll highlight a couple anyway. Obviously you make it easy for your customers to get in there. It's template based environment for large projects organizations which I think is amazing. You have dynamic resource allocation depending on where the needs are. It might be geographic like you had mentioned in the USA if they've got a higher demand you can easily shift that. Your AI sandbox obviously makes it easy to stand up and then deploy. I really like that. And then offering those AI APIs that you can easily plug in and get right to work. So I appreciate that. Pratik, it was such a pleasure having you on today and I look forward to speaking with you real soon.
- [14:13]
B
Thank you. Thank you for inviting. It was great to be on podcast.
- [14:17]
A
Bye for now.