Good morning everyone. My name is, and I'm a resource driven IT professional. With over 12 years of expertise in software industry, my career has been dedicated to ensuring the reliability, performance, and compliance of enterprise applications across critical industries, including telecommunications and healthcare. Currently as a lead technician at Spectrum, I swear, head software initiatives for billing systems. Validate so plus a PS and design UI UX workforce, leveraging tools like Postman sql, no SQL databases and automation frameworks. My work directly supports CICD pipeline and Agile in ensuring seamless integration and high impact solutions. Previously. Alert software efforts for Medicaid chip systems under Department of Health and Human Services for South Carolina and IBM Vera Ensure HIPAA compliance and federal data hub integrations. Demonstrating my depth and regularly demands with proficiency in automation, defect tracking, and cross-function collaborations. I bridge technical and business needs to deliver robust software solutions to my portfolio includes 200 plus technical documentation and thousand plus automation revision suites and scalable testing strategies that reduce the risk and accelerate delivery accelerate delivery. And I had a master's degree from the University of New Haven, and I'm passionate about mentoring teams and driving driving innovation in software practices. I'm excited to share my insights from experience and explore how we can solve complex challenges together. This is the this is about building rust power AI for telecom from reactive to predictive systems. Today as we are driving into transformative shift into telecommunications, we are rush forward. AI is turning reactive network management into proactive predictive systems. Imagine a world where outer is separated and prevented before even customer notice. This isn't science fiction. It's happening right now. The rust is the answer here to make it possible. Where cutting gate technologies meets real world impact. Let's explore how revolutionary is unfolding it. Yeah, today's agenda is how he's our today's agenda, the telecom challenge. Why today's break fix model is broken. The second one, the rust. How a language designed for safety and speed solves decades, all problems. The real world implementations from anomaly detection to predicting cell tower failures, and now about the integration challenges, the gritty reality of merging AI with legacy systems. The fifth one is future roadmap. How to face this into operations. Why does this even matter? Because every minute of downtime costs silicon communication on average of $5,600 per minute. The stakes are high and the solution is here. The telecom challenges today. So let's confirm the harsh reality of today's networks. 72 percent's fault detection means 28% of the issues slip through silently degrading the services of the telecommunication, sir, 3.7 hours for MTTR. Picture a hospital losing connectivity during an emergency every minute counts. 80% of today's are planned outages where like mechanics changing on a schedule where the engine is already overheating all kinds of things, which is the root cause, the fragmented tools. Engineers today tools are between 80 years to 12, 80 to 12. Systems wasting 20 plus hours weekly on manual correlation reactive mindset alerts, fire after customers complying. I worked with a European carrier last year who discovered that 40% of their sudden outages had clear procedures like memory leaks or temperature spikes visible in the tele data days in advance. But without ai, no one connected the dots. Now the cost is that's the cost of reactivity. Why rust for telecommunication ai? Rust is in just another language. It's a paradigm shift. Let's break down why it is a perfect for telecommunication. We have the memory safety garbage collectors in Java. Python causes latency spikes like 200 milliseconds. Pauses us in telecommunication industry thus enough to miss a critical failure signal. The RUS ownership model elements, the elements. This one, the real world example, is a tire one. Carrier company reduced elemental processing from one 50 milliseconds to millisecond to milliseconds by switching it to rust. The features of concurrency, traditional threading leads to race conditions. Rust Comply forces threat safety. A case study saying that the processing 10 million data points per second across 50 threats with zero deadlocks, the zero cost attractions for this particular thing. CC plus risk level speeding without crashes. But benchmark of the rust process upward. Hundred GB of network lags in 22 seconds versus python's eight minutes. That is where the real difference is the strong type system. The compile time checks are caught. A caught a critical API mismatch in a billing system integration that would cost a 2 million in NRS changes. A fun fact about this whole thing is. Most love language on Stack Overflow has been for over the seven, eight years. Why? Because it's runtime disasters in compiled time errors for telecommunication. That's the difference between a hiccup and a headline marking making outages. This is about the performance metrics, whether the traditional implementation let's talk the numbers. The slide compares rust power, AI to the legacy systems. The fall detection 92% was a 72% catching. There's 20% more issues means fewer midnight emergencies. And PTR 47 minutes was a 3.7 hours. That's a three hour save per incident. The three hour save per incident, customer searching down by 1.2%, which translate to $8 million and a year for a mid-size carrier. One client said, why not just scale our existing Python system as we tested it? It's 5,000 requests per second, their Python service cluster, but the rust handle 85,000 requests per second on the same hardware. Sometime the tool is bottleneck. The last powered AI algorithms. This is about this thing. The two algorithms are the game changers. The first one is isolation forests. How it works is randomly isolates. Anomal is, instead of profiling normal behavior think is it's like a folding haystack and elevating hay. It's detecting as A-D-D-O-S attacks in, in three seconds by spotting irregular traffic patterns, LSTs, these neural networks predict failures. By remembering patterns over time. So this is the main, key thing in this whole thing. An example, a cell tower. Power supplies typically de degrades over six months. The LSTM flag once failing for every two weeks, saving almost a 200 K in replacement costs. See the real fear and the whole kicker is the rust speeds. Let us run these models in line with data streams. No batch crossing Python would need 15 times more servers to keep up with this 500 K in year on this cloud as its cost has been saved, particularly using this rust techno rust language. Thus this slide seven case a case study about predictive maintenance of the cell tower led zoom into North American carrier results the 25 tb. Per day of data from 5,000 cell tower cell towers eco into 10 years of Netflix streaming the LSDM predictions, the flagged falling hardware 14 days in advance with 89% accuracy results, which in turns result to 60% fewer outages and 42% fewer truckloads saving on an average of $1.2 million in a year. This is the best part where the NOCT went from firefighting to strategic planning. One engineer told me that finally feel like I'm preventing disasters, not layering them up. That's a cultural transformation in this telecom industry. The next slide slows about the unified pla platform architecture in this particular thing. So here is how the sausage. Gets mo gets more, gets made. So collection adapters, Russ normalize the data from eight to 12 legacy systems into a unified format. Then the message bus, the message base zero copied design moves a five B in an hour with the overhead of pass post-it note. So explainability layer. Where the engineers get plain English insights like Transformer 42 fails 70% of the time during peak loads. Now we can notice that in 1990, Sarahs a Sonet system or rust, FFI, the foreign function interface. Lets let us wrap. Its c code safety. No re rewrite needed. The client called it a time machine for their stacks. That's back in 1990s. Era. This are implementation challenges. So no transformation in painless, so it's always integration Challenges are already there. Here, how we can tackle this hurdles. So the legacy integration, so use plus FFI to wrap COBOL billing systems, there's a tip there. A first testing caught a memory leak that would crash production. And next is the data quality. Rest scripts autofill, missing timestamps using Marco chains. And the next one is the black box sphere added to y button next to every a alert. That's how it is showing the top three contributing factors. A CTO once asked me how could I sell this to my board? The answers straightforward. The straightforward, start with the pilot on a non-critical towards. When they saw 40% reduction in outages, the budget approval took 10 minutes. It just took 10 minutes. No. The next one is about the workflow evolution of the operational rules. So AI won't replace any engineers. It redefines the roles before the manual log scripting for the outage triage after this one, the 20 motors and turning algorithms. Once a team automated 80% of the repetitive task instead of layoffs they upskill everyone. Their job satisfaction score jumped 30%. Happiness engineers fix problems faster than the regular one. The next slide is about the implementation roadmap. Of particularly, there are four steps in this implementation. The phase one is about the data foundation. Deploy rust collectors. It's a tip in instrument like one tower type, first example 4G only. The phase two is anomaly detection. Use isolation for us to find quick wins. Its example is a power supplies, a faulty power supplies. The phase three is predictive models, training LS teams on historical failures. Caution, start with a non-safety critical system, sir, and coming to the phase four is automation. Auto restart. Failed notes with, without an even human oversight client. As an example, a client faced this over 18 months by six 18 months, but by six, the ROI was 2.3 million from reduced truck roll alone. The boat the fast truck, the whole thing the whole thing. The next one is thank you. Good.