Cognitive Shield Real Time Real Smart - Rachna Srivastava

Hello everyone, thanks so much for joining me today. Let me start by asking you a very simple but profound question: what happened when the smartest tool we've ever built stopped working for us and started working against us? Let's just stop for a moment and think about it.

Imagine getting a phone call: a voice sounds exactly like your manager, the same tone, theurgency, the choice of word, everything on the spot. He says his intranet is down and asks you tourgently send all the confidential information to his personal email so that he can brief a client.

It all seems legitimate, until you later discover it was a deepfake scam. You unknowingly sent company-sensitive information to a frosture. Or picture this: a face appears on your screen for a video KYC, blinks at theright time, smiles naturally, clears identification without a hedge.

But here's the catch: it's not real, it's AI-generated deepfake. We all know it already. So the issue is, we're not dealing with old-school fraud anymore; we're facing synthetic identities, deepfake onboarding, and AI-driven scams that look and feel more human than human itself.

These threats do not just break in; they get verified and walk through the front door completely undetected. The issue is, we're not just trying to detect fraud now; we're trying to detect intelligence. And that's the real challenge, and it is exactly why we need to rethink everything we know about trust, identity, and defense in the age of AI.

So how do we fight back? How do we stay ahead in the world where fraud does not just hide, it blends in?

This is where Cognitive Shields comes in. Today I will show you how our cutting-edge solution leverages advanced AI and machine learning to spot and stop these invisible threats. But before we dive in, let's start with some real stories.

Here are 3 real-life stories that highlight how AI-driven fraud is affecting people today. The first story is a story of Anthony, and this story is about a voice cloning scam. Let's listen to it. Anthony is a retired father, lives in California.

One afternoon he got a phone call, the voice on the line undeniably his son's. Same accent, same little tone, only a father would recognize. The son sounded panicked. There has been a terrible accident, a pregnant woman was hurt, and he was at a police station and needed bail money immediately.

A moment later, another man called, and he claimed to be his son's lawyer, and heurged Anthony to wire $50,000 immediately or his son's would be taken to jail.

Anthony has never heard about deepfake, but he had heard about his son's voice and he trusted it. He immediately wired $50,000, his entire retirement savings, later to know it wasn't his son. It was a AI-generated voice clone created using publicly available TikTok video of his son.

This time, by the time the real son came in the evening, the money was already gone.

Let's listen to the second story, and this story is about Lisa, a 45-year-old woman who lives in Ohio. She feels very isolated, more isolated after the pandemic. One night, a man messaged her on her Instagram, claiming to be a famous Australian TV star.

He called her his soulmate and promised to marry her. And this continues for over 18 months. They message every day, but never met, always blaming visa and money issues. He asked for help, and Lisa sent nearly $40,000 of her savings over time.

But the man wasn't real. His face was made by AI, and this was a scam. These scams actually have a name, and this is called pig butchering, and they're growing very fast. Scammers in this case build fake relationships to steal money, often using AI and crypto to hide their track.

And Lisa reported it in January 2025, and she shared her story to warn others, and she feels if it is too perfect to be true, it might be a scam. Let's talk about the third and the last story, and this is the story of Xavier.

Xavier is a 29-year-old accountant from Austin, Texas. He's really smart, financially savvy, and always on a look for a real big tech opportunity. He thought he found one in early 2025, and that's when he discovered Zipmax Pro, a flashy new cryptocurrency project that seems to check all theright boxes.

It had everything: a very slick, professional-looking website, dozens of growing investor testimonials on YouTube, a white paper filled with current cutting-edge AI and blockchain jargon, an active Discord channel run by charismatic developers, weekly live streams, and AMAs featuring synthetic avatars modeled after real Silicon Valley influencers, and even a deepfake video of Elon Musk appearing to endorse the project.

And the pitch was very simple: an AI-driven platform that optimizes DeFi investment and promises up to 35% annual return. Xavier, like thousands of others, believed he was getting in very early to this next big thing. He invested $60,000 of his personal savings and an entire 401(k).

Then, without warning, it all disappeared one day. The creator of Zipmax Pro executed a classic rug pull, dumping their holdings, crashing the coin value to a plummet.

Xavier lost everything, but he wasn't alone; over 5,000 people across the U.S. were defrauded by the same scam. And the worst part? Every element of the scam was powered by AI: fake ID verification to pass the crypto exchange check, deepfake celebrity endorsements, AI-written smart contracts that looked legitimate, social media bought as synthetic influencers to build hype.

This wasn't just a isolated incident. AI-powered scams have surged 375% since 2023. 76% of synthetic identities now bypass traditional fraud detection. Americans reported record $9.3 billion in losses from crypto-related crime, and this is a 66% jump in just one year.

These are not phishing emails of the past. They are intelligent, emotionally engineered attacks built by machines and designed to exploit trust at scale.

As AI continues to evolve, so do the tools of frostures. So it's imperative that we develop a robust defense to protect individuals from such sophisticated scams.

So now, let's be honest: AI can be used to deceive, to defraud, to exploit. But here is the good news: AI can also be used to detect, to defend, and protect. And that's the paradox we're living in. It is the one we have to embrace.

The same AI that is used to commit fraud can be trained to stop fraud. The same model designed to manipulate behavior can be retrained to recognize it and shut it down. The same technology that is shaking our foundation of trust, it can be reused to rebuild trust, and it's stronger than ever.

And in this presentation, I'm going to show you exactly how.

So now that we have laid out the challenges and the high stakes of AI-driven fraud, it's time to talk about the solution. Let me introduce you to Cognitive Shield, the next-generation platform designed specifically to protect the financial ecosystem against these sophisticated threats.

Cognitive Shield is designed as a simple three-layer defense system. Each layer is tackling a different part of the fraud problem, from prevention of real-time detection all the way to intelligence response.

The layer one is all about building a strong foundation. This is where we securely manage user data, licensing data, examination cases, and payment data. But it's more than just a storage layer. We use AI to guide users through the complex processes and flag potential risk before they become a real problem.

It is safe, it is smart, and it is user-friendly. Next comes the layer two. And layer two is about real-time fraud detection engine. This is where AI really shines. Our system uses 8 advanced detection modules, constantly scanning for threats from deepfake bots, phishing attacks, synthetic identities, crypto scams, and more.

We also use graph technology to map user and transactional behavior, helping to spot the fraud rings that traditional systems usually miss. Finally, layer three brings all the things together by combining AI and human insights for a smarter response.

Investigators get powerful consoles with AI-powered search and live dashboards and trend analysis. Cases are automatically escalated when needed, and the system keeps a complete audit trail of compliance. This means teams can move faster, stay organized, and meet regulatory demand effortlessly.

In short, Cognitive Shield brings together smart, secure data management, real-time AI detection, and intelligent human-led responses, all in one platform, so that the organization can stay ahead of fraud every step of the way.

Let's start with layer one, the foundation of Cognitive Shield, our secure user and regulatory management layer. This is where all the core operations happen: the licensing, application examination, case tracking, payment, etc. It is built on a secure, reliable database so that your data stays protected and organized.

But what really sets this layer apart is how deeply AI is integrated into every step of the process. For example, when a user submits a licensed application or

renewal application, AI instantly checks for missing information, flags inconsistencies, and offers real-time guidance. It is like having an expert watching over every form and making the processes smoother and error-free. When agencies launch an exam, AI reviewers respond and document to spot unusual patterns, potential red flags, helping teams focus only where human attention is really needed.

On the legal and billing side, AI breaks down the complex cases, outcomes, clarifies fines and deadlines, answers any user questions in plain and everyday language. No more digging into

legal jargon anymore. It also has a smart built-in assistant that users can use to ask questions naturally, upload legal documents, get quick summaries, insights, all in one place. And finally, everything is presented in a role-specific dashboard. Whether you are a regulator, a licenser, or an auditor, you get a clear view of the application, compliance status, and payment workflow.

So layer one isn't about managing data; it's about turning the complex processes into a seamless, intelligent experience.

Now, let's delve into layer two, the core of Cognitive Shield's system. This layer is engineered to identify and mitigate sophisticated fraud attempts in real time, leveraging state-of-the-art AI technologies.

Our system comprises 8 specialized detection modules, each tailored to identify a specific type of fraudulent activity. Deepfake detection, for example, utilizes a generative adversarial network, a GAN-based system, to identify and manipulate media. Bot detection, for example, employs a machine learning classifier, a gradient boosting machine, to discern automated bot activities in blockchain transactions.

Phishing detection analyzes the communication pattern using natural language processing to detect AI-generated phishing attempts, using WHOIS and other techniques. Crypto scam generation applies a graph neural network to analyze transaction networks, identify anomaly patterns in the fraudulent activities. To power our fraud detection engine, we use some of the most advanced AI technologies that are built to understand and respond in real time.

For example, we use deep learning that helps us analyze images and audio to detect things like deepfake voice cloning quickly and accurately. Graph neural networks track connections between users, devices, and transactions, spotting hidden fraud rings and suspicious patterns that we'd miss otherwise.

Natural language processing reads and interprets text to detect phishing attempts, social engineering tricks, unusual language in communication, etc. And we finally have multimodal signal processing that pulls all together in the text and voice metadata. We get a full picture of the thread, and we can respond smartly.

Let's talk about how we use graph-powered AI to find hidden fraud.

Fraud is not always one bad actor. It is often a network of connected people, accounts, and devices, so that we have to focus on how things are connected, not just what happened. And here is how we do it in three simple steps.

Step one is building the graph. One of the hardest parts of graph-based fraud detection is turning the unstructured data into a structured, graphical knowledge base. And we solve this using an agentic workflow that we built using CrewAI and a large language model.

Here, we can

extract the entities and relationships from simple text: PDF documents, forms, emails, logs, etc. We also enrich these graphs with information from the internal Postgres database that helps us tie everything together to create the complete real-time view of the fraud landscape.

Then we run models like GNN to find the hidden connections, things like groups of accounts that are acting in sync, devices that have been reused across multiple fake identities, and so on. In short, we have automated the most difficult part of the graph intelligence, that is creating the knowledge graph, and we turned it into a powerful tool for uncovering fraud rings that would otherwise stay hidden.

Step two is Neo4j is used as a graph persistence mechanism. We store all the graphs and notes and relationships into a Neo4j open graph database. And step three is asking graph-smart questions. And this we do by using a Neo4j-based retrieval augmented generation, or RAG system, that is integrated with a large language model to take the user queries in natural language and then translate that into a ciphered language that is understood by Neo4j.

And

that allows us to generate

user

queries

seamlessly by taking them in the natural language. This setup enables real-time exploitation of graph relationships, supporting high-performance fraud detection by surfacing patterns, anomalies, and entities linkages that traditional relational systems often underlook.

So let's talk about layer three, and this is where everything comes together. As fraud becomes more advanced, our responses need to be smarter, faster, and more coordinated. This layer is all about that, turning alerts into action and action into results.

So first step is the unified fraud intelligence console. Think of this as a mission control. It brings the insights from across the system into one place. And the best part is it uses AI-powered natural language search to investigate so that users do not need to remember complex queries to get the insights from the data, whether it is a Postgres database or Neo4j database.

Second step is real-time dashboards and adaptive analytics. This dashboard gives us the live view of what is happening: fraud hotspots, trending tactics, big actors are connected.

It's where you will see the visual intelligence that helps

teams move faster and make more informed decisions by looking into these dashboards in real time.

Then we have a case escalation system, or an alerting system. And what it does, it

spots the serious threat. It does not just flag it; it acts on it also. AI automatically analyzes how severe the open case is, routes it to theright person or the team. It uses the mix of rule-based and LLM-based logic to decide what needs more attention and when.

And everything is logged with role-based access and a full audit trail. Finally, we have compliance-ready reporting. All investigations are fully traceable. Reports can be exported in PDF or in CSV form. And this helps regulators, auditors, and the internal team.

Everything is clear, well-documented, and easy to share. So in short, layer three is where insights become action. It helps you detect, respond, escalate, and report, all in real time and all with complete transparency.

Before we get into the architecture of the system, let's stop here and let's see how. Now you have the background. Let's see how the real system looks like.

I will give you a walkthrough of this is the Cognitive Shield application. Here you see the Cognitive Shield. This is the main dashboard where you see what are the recent searches, what are the emerging threads, what are the recent alerts.

And as we mentioned before, this is a three-layer system. Layer one is about user management. Layer two is about fraud detection and also about graph fraud detection. And layer three is about investigations and search and dashboarding, alerting, and so on.

So we start with user management. This is where the user activities are recorded. Security settings are done, user account management, user contact management. Everything related to the user is performed in this flow. Then we have the case management.

In the case management, you can create a case. You can analyze the case using AI. You can search for cases and so on. Then we have examination management. Here, any cases, you can look at the scheduled exam, all in-progress exam, completed exam, canceled exam.

You can create a chart examination on any cases. You can manage existing exams. You can perform risk assessment, risk history, business rules. You can also view the timing of the system. You can view the expense report, billing thing analysis, all kinds of analytics and reporting.

Then we have the invoice management system. Every system, as you see, all flows start with a dashboard where you see the summary of what you have done. Here you have invoices that are draft, sent, paid, overdue, canceled. You can create a new invoice.

You can export the invoice. You can get the AI analysis of the invoice. Then we have a processing flow. Here you can link different organizations. You can

perform the calculation. You can get the discrepancy detection using AI. We have a payment portal. In this payment portal, you can see the pending payment, payment history, different kinds of receipts, and so on. Then we have a chat assistant.

This is a multimodal chat assistant that is based on voice output, voice input, web search, different modalities of search. Here you can browse and save user data that you want to search on, and so on. Then we have a different kind of fraud detection system.

You can perform a deepfake fraud detection. And in this case, you can perform a deepfake fraud on images, videos, audios. And you can get the recent analysis. You can search by different deepfake fraud. There is threat intelligence and so on.

Then we have payment fraud. What kind of suspicious transactions are observed? What kind of active cases? What kind of a response time? You can see the fraud trend over time, fraud by different types, alert statistics, also the payment analysis using AI, fraud patterns observed, different case management.

Then we have web fraud. These are the real-time fraud monitoring. Then we have social media fraud. Here you can choose the different social media from where you different fraud type, the investment scam, phishing scam, and crypto scam, and so on, observed on different social media.

Then we have cryptocurrency fraud. You can get the suspicious crypto wallet, smart contract, transactional flow. You can watch the suspicious activity in blockchain, financial impact by different blockchain, cryptocurrency schemes, fraud trend, and so on. Then we have phishing detection.

What are the different active phishing campaigns going on? Domain analysis. What are the targeted organizations? And then we have bot detection. What are the different bot networks, bot behaviors, threat analytics, and so on. Similar, we also have graph-based fraud detection, knowledge graph builder.

Here you connect to the Neo4j. We have text-to-graph. It helps you create the knowledge graph based on the user text. You can create the graph based on the PDF document or

document file. And then we have a query generator that can generate a ciphered query based on your user natural language. And we have a graph RAG to extract the query and get the insights.

Similarly, we have advanced

investigation and enforcement. This is where users can perform the searches, general searches, domain regulation monitoring, and you can also have a search history. You have analytics dashboard of the whole system. And also you can get the AI insights from also the transaction and alerting system, fraud pattern, geographic distribution of the threats, risk model performance.

We have alert monitoring. Here you can choose the severity of the attack, the status of attack, and the sources where you want to filter the attacks and get the dashboard. It also saves your attacks and statistics. And you can perform the active investigations on different frauds and so on.

So this is basically a very high-level what we

have done. And we also have different here are the APIs that we have designed to

it's an API-driven architecture where APIs are designed and used to extract the data and manage

and extract the data and manage from the front end. So coming back to our system. So let's learn about how we have actually built this

system. Cognitive Shield is just not another fraud tool. It's a real smart AI-enabled, AI-supported tool that handles the modern fraud from deepfake to crypto scam to social engineering and so on. The front end we have built using Streamlit to create easy-to-use real-time dashboards.

API layer is built using FastAPI that handles all the incoming data, whether it is the login, transaction, document upload, and so on. AI layer is powered by CrewAI. That is the brain of the system. It runs multiple AI agents that work together and get the insights.

The data layer is done in Postgres database. Neo4j is used for graph analysis. And then we have Graph RAG and LangChain for AI agents.

So the whole system is designed to build and scale a real-time detection of the different frauds. This is AI fraud defense, which is smart, fast, and built for today's threats.

Let's

talk about what we have really learned after building this system. We learned that if you build a system, we have to start with security from day one. Trust is not something you patch in later. It has to be ingrained into the system.

Next, do not rely on a single AI model to catch all kinds of fraud or all kinds of fraud. Fraud is a messy, fast-changing, so one size is not going to fit all kinds of fraud problems. So we have to use multiple specialized agents, each one turned from a specific task, and let them collaborate in an agentic manner.

Then always think in graph, not just rows and columns of a relational database. Graph help us detect the hidden connections, which we usually miss in the relational database.

And instead of building a giant monolithic system, think of microservices, and we use FastAPI, API-driven architecture that can be easily scaled. And we have to keep observability in mind. We have to monitor the AI models.

We have to monitor the uptime. We have to monitor the false positive, false negative. We have to track everything. And we have to make sure that every decision is explainable. And that is how we earn trust. Finally, build the privacy from the start.

We encrypted everything. We assumed nothing. And with these principles, we made Cognitive Shield a resilient, transparent, and built for the real world of financial fraud.

As we close today's presentation, let's try to summarize

and bring everything together.

We have learned that Cognitive Shield is built by a three-step system. Step one is about data trust, data storage, and everything starts by keeping the privacy of the user, user security, and also using AI to help the user manage and

store the data. Step two is real-time fraud detection. This is the brain in action, powered by deep learning, Graph AI, secure LLM, that do not just react to fraud. They anticipate it and work on it.

Step three is intelligence hub for responses and compliance. This is where the insights become action. We surface the signals, the alerts, the visual dashboard, and compliance reporting.

Here is what the key takeaway is that AI is not an optional tool. It is the future

of fraud defense. Graph over tables because it's essential, because relationship revealed by relational database has to be also supported by graph to capture the network of connections.

Multi-agent LLMs give us the speed, clarity, and context of the world where milliseconds matter.

Why do we have to act now? Because if we wait, by 2027, 90% of cyber attacks will be AI-driven. Fraud losses will surpass $100 billion per year.

So we cannot wait. Because our mission is very clear. We have to stop fraud before it starts. This is not just a platform. It is just a technical it is a movement. And we want you to be part of it.

Let's defend trust. Let's protect the future. And let's do it all together.

As we wrap up this presentation, I want to take a moment to dedicate this presentation to someone who has been a true inspiration in my journey, Jeremy Howard, the visionary behind Fast.AI. He did not just teach me how to build models.

Everything I know about deep learning, I learned from his classes, his code, his belief that AI should be open, ethical, and accessible to all. His work changed the field of AI.

ULM Fit redefined how we approach NLP. But more important, his teaching changed people, including myself. This platform, this mission, this entire journey is built on what you gave us all. Thank you so much, Jeremy, for your vision, generosity, courage, and continue to inspire us all.

Thank you so much, everybody. Thanks.