Good morning, good evening, good afternoon, wherever you are. Thank you for joining the session. I am goes by Sid. I'm going to talk about building a high performance multi-agent AI system in Ross. How do we use Ross to build a multi-agent system and how it can achieve 40% faster enterprise transformation using its memory safety guarantee? Let's deep dive into it. Sure. The enterprise challenge today, every digital transformation passes critical bottleneck. The 70 per 73% of organization struggle with the monolithic AI system, creating a single point of failure. Everyone they want to build AI system, AI platform, but they face a single point of failure because the design itself is wrong. The traditional centralized architecture is definitely a garbage collected language. Suffering from an unpredictable latency spike. A memory of overhead can deploy processing up to 300% compared to RO based distributed approach. So these are the basic critical bottleneck people generally doesn't realize. They're going to face a challenge in their digital transformation and, and in enterprise AI challenges together. But advantage of cost is, for a multi-agent system is it is faster in process optimization. It can reduce your risk. It has a greater efficiency and has a zero memory leak. Compared to traditional language and framework, Ross is 40% faster and the lower implementation failure through compile times safety, 60%. And it's improve the throughput. Aging Roth concurrency, which is like a great efficiency you get, which is like 80%, 85% more than the other language we used. It is always a zero memory leak. Roth ownership model and zero cost obstruction enables a back through multi-agent framework and that deliver a measurable enterprise advantages. Let's talk about a little bit into the. Architecture. The distributed architecture using Ross, it has a four different layer. The first layer is the perception layer, then cognitive layer, action layer, coordination layer. The perception layer is specialized agent from a data eson. It is using for filtering, normalization, using the RO efficient ai preemptives. The cognitive layer is like little bit of intelligent analytical layer, which use lost. Processing for modern inference and decision logic. The external layer is an education of agent. It's really what it does and in implementing your business logic in the transactional flow and it's get you the guarantee of the excellence. The coordination layer is mostly the orchestration of the managing different workflow in an across distributed system and boundary. The key are. Using, using ROS are mostly the memory, safety, concurrency, performance, reliability, and interoperability. The memory setti is the biggest deal using ros. You like. There's a zero memory leak across 50 plus million agent interacting using RO or system concurrency by using a Tokyo best S Sync on time. It can handling 250% more concurrent agent than it's equivalent Python, or go on the performance. 40% reduction in workflow education time throughout Ross. Zero cost abstraction reliability. It's 60% decrease in system failure impact, lost compile time error prevention, interoperability. P zero three PI oh three and was binding enables lost. Its in interoperability with any legacy system and its ability to do that is close to 80% successful than any other language provided in the market. The memory set in multi-agent system, which is critical for you to build any kind of multi-agent distributed system. The challenge is a complex oxy pattern of agent of interaction, cascading failure due to memory leak and garbage collection, POS disruption, realtime coordination among the system. Roster is going to help you on this. The Ross solution is, it's. The model and compiled and guarantee eliminate the entire memory related bug crucial for your high performance and real time multi-agent system. The practical benefit you're going to get are consistency in performance, resource efficiency, and reduction. Bugging our production system. Demonstrate the zero zero memory leak across 50 million plus agent interaction and high throughput and propag environment. The concurrency in Tokyo. West Agent Coordination. So this is again, one of the platform Ross provide you to use. It's an asy on time. It enables high, concurrent, and efficient agent coordination through an async environment. This approach drastically reduce your OV rate overhead and boost responsible multi-agent system, maximize throughput agent performance. And then it's help you in my patient resource utilization. It's enhance responsiveness is help you in robustness. Our system is achieved generally 250% more concurrent adjunct education than it's equivalent to Python or go and any, any other language used for it. So this is this Tokyo, this, this platform after us like Tokyo compound its impact in the real world performance in terms of a multi, multi-agent distributor system. Performance is another aspect of it zero cost obstruction. If you see the last performance versus the traditional one, it's, it's like 40% reduction in workflow time. Then any other competitor compiled time optimization, it's eliminate your on time checks. No garbage collection. Pause during critical operation agent communicate overhead reduced by 60% because of this architecture. And this implementation has spread pretty easy because it's a custom trade. Zero overhead agent polymorphism static dispersed, where agents types are known, efficient serialization for intelligent communication automatically. It's built on this architecture by ro. And then kiros ecosystem and tools. As I said we used the Tokyo based, I'm going to talk about a little bit more on that. And, but you have other, other different type of architecture and platform. You can definitely use it. We used this Tokyo one, the first one. It's a, a synchron on time for an efficient agent, scheduling and coordination. It enables non-blocking your AI across boundaries. And then it enables priority based on that. Support complex workflow and communication. So and then similarly you have Actos and Sled and then Kendall, different type of interfaces and platform to think about, system preventing distributed system bug, like any, any multi-agent distributor system. You talk about it, there is a common enterprise AI failure. Either it could be a data loss condition, it could be a parallel state update. It could be inconsistent error handling or a protocol violation. Rust help you to overcome this because it's a inbuilt compiler help you to catches these errors. 80% chances, 78 to 87%. Chances this box will be a catch before your deployment compared to Python or go any other languages you use. So this is one of the greatest thing you got from rust. This is another case study. And in the manufacturing company we have built the challenge was it's a global manufacturing company struggled with an unpredictable latency in quality control AI system. Leading to 12% is a production delay. It is a big deal for them. When it's a multi billion company with an MNC you do, this is a big problem. The result was like 43% reduction in high quality inspection time with zero increase in defect. Right. And also the latency was less than 50 millisecond with the three X normal load. And you give three x time of the actual load still, the LA was. Less than 50 millisecond, which is the outcome of this whole architecture after implementing it. Soroto multi-agent solution. It is, it is, it is. It is like we, we built around 22 specialized agent distributed across four architecture layer. The one we we just signed, and the real time coordination inspections and decision and routing for different functions. And the integration among those systems, and also it bell integrated with the legacy system of the company using UI oh three and web centric pages. Which is a great aspect of rust. The implementation roadmap generally took it takes a 20 week, a minimum depending on what, what you want to do and know how big or how small it is starting from your architecture assessment to core infrastructure to, and is on integration with existing enterprise system and production deployment and optimization. Then the available resources are like complete rust implementation performance benchmark and deployment strategy and integration pattern. This has to be your main resource to thought about and, you know go over it. People, even you think about implementing it. Thank you for joining this session. Hopefully this will help you. Thank you very much.