Hello everyone. Pratish and I are excited to welcome you to our talk on healthcare chatbots and explainable AI. Today we'll delve into the evolving role of healthcare chatbots, the integration of generative AI capabilities, and the critical importance of explainable AI. We'll explore their current impact, future potential, and how they are transforming the healthcare landscape. Towards the end. We'll also discuss the challenges these technologies face and potential strategies for their broader adoption. Today's agenda is as displayed on the screen. We hope you'll enjoy this talk. Over to Radhesh. Thanks, Abhishek. Let's begin with how AI powered chatbots are transforming healthcare by addressing accessibility, cost, and personalization challenges. AI chatbots alleviate geographical and scheduling barriers, ensuring 24 7 support for patients. They provide personalized care by adapting to individual patient data, enhancing engagement. And they also streamline processes like symptom checking and chronic disease management, making healthcare more efficient. These capabilities set the stage for how chatbots revolutionize healthcare delivery. Expanding on the need for 24X's ability, one of the primary benefits of AI chatbots is their ability to provide round the clock support. Unlike traditional systems, limited by office hours, chatbots are always available to guide patients or escalate care when necessary. They also support chronic condition management through continuous monitoring and reminders. Triage ing symptoms quickly saves time and provides healthcare providers with tools they can use to identify care. A great example could be, imagine a diabetic patient receiving a reminder to check their blood sugar if there is a glucose spike or a drop, so that the chatbot can ensure the signal can be captured for appropriate treatment. This brings us to how Genitive AI is changing healthcare chatbots. Genitive AI takes healthcare chatbots to the next level by making interactions more dynamic and patient specific. These chatbots deliver contextual responses, enhancing the user experience. They simplify medical jargon into patient friendly language, improving clarity and engagement. This personalization builds trust and ensures patients feel understood. One of the key elements in chatbots is their ability to retain what has been already discussed. Has it happened many times to us that repeating ourselves to different doctors tend to frustrate us over and over again? AI chatbots, due to their vast context retention, aim to eliminate that and have a coherent patient history while talking to them, remembering most of the things that they have asked the chatbots to remember or stress the importance on. Context retention ensures chatbots remember past interaction and provide continuity. This enhances trust, reduces frustrations, and also speeds up support. Another key element. In chatbots is the amount of memory the chatbots can retain. AI chatbots don't just assist patients, they also evolve over time. Using deep learning and reinforcement learning, they refine their interactions for better accuracy. Technologies like Knowledge Graph help them contextualize medical data effectively. This continuous learning ensures the chatbot improves with every interaction. This brings us to one of the core topics of our discussion today, which is explainable AI. In healthcare, blind trust in AI isn't enough. Patients need to understand why and how decisions about their health are being made. While the benefits are immense, we must also acknowledge that we need to address the ethical challenges. Explainable AI intends to transform the black box into transparent medical assistance. Transparency ensures patients know that they are interacting with AI. Healthcare professionals also need to verify the logic used for the decision. or the recommendations made by the healthcare jackpot. Healthcare AI can also get help from explainable AI to meet regulatory and compliance standards set by governments and institutions. To the patients, it's also beneficial because it can, again, just like explaining it to the healthcare providers, give tools to validate the logic while remaining in compliance. Let us now look at how we can actually achieve this transparency through some of the specific explainable AI methods. Explainable AI methods broadly fall into two categories, the model agnostic ones and the model specific ones. Some of the model agnostic ones are discussed here, such as line, Okay. which tries to explain responses through approximations useful for explaining complex decision making. SHAP, where each input feature gets an importance value, or the counterfactual method, which is a classic way of assessing the what if scenarios. All of these methods have been widely used across machine learning and AI systems to interpret and explain ML models. Let's now look at some of the more model specific methods. Some of these methods might seem familiar to those who come from a machine learning background, such as decision trees. or feature importance analysis. They have a lot of things in common where decision trees, just like in traditional methods, can create feature importances that can be interpreted for future use. Model specific methods can help understand how each feature contributes to the outcome and can be communicated to the healthcare professionals providing that care. Feature analysis uses each input feature and assigns an importance value to it. Whereas layer wise propagations, such as used in clinical trials, The convolution neural networks or traditional neural networks can help give the same information. I'll hand it over to Abhishek now to talk about more on the data engineering side of things. Thank you, Rathesh. Behind every effective AI chatbot is a solid data engineering framework. Robust data engineering ensures clean, reliable, and scalable pipelines that drive chatbot algorithms. These pipelines enable real time data ingestion, storage, and processing, which is critical for immediate and accurate responses. Seamless integration of diverse data sources to support personalized and context aware interactions. The key components include ETL pipelines, data lakes to store diverse data sets, both structured and unstructured, such as medical images, patient notes, and device readings. Also, data warehouses enable fast querying and reporting, helping healthcare providers quickly retrieve insights for decision making. Now let's dive into the architecture that powers these chatbots. The workflow consists of several interconnected components, starting with data sources, which store patient records, IoT devices like wearables, and medical knowledge graphs from the backbone of the data feeding into chatbots, batch processing, such as daily updates, ensuring data consistency, while real time streaming also allows chatbots to respond instantly to events like a patient's abnormal heart rate. After data pipeline is set up, we use the data for AI model training, enabling chatbots to deliver personalized and mathematically sound interactions. Finally, chatbots integrate with mobile apps or web platforms, offering patients easy access anytime, anywhere. While these technologies offer incredible potential, they also raise important ethical questions. Given the sensitivity of Healthcare data, secure storage and secure transmission are paramount to maintaining patient trust. Patients must know they're interacting with AI and give informed consent for how their data is used. AI models must be trained on diverse data sets to avoid biases that could lead to inaccurate or unsafe recommendations. AI should complement, not replace, human healthcare providers. The goal is to enhance care while retaining the human touch. Now, I will hand it back to Ratesh for closing comments. Thanks, Abhishek. Of course, no technology is without its challenges. Running AI systems is already a complex task, and add to it explainable AI and the requirement goes even higher. We need to address technical reliability, operational training, and ethical guidelines likewise. Clear regulations on privacy and consent are essential for widespread adoption. Guidelines like HIPAA needs to be followed strictly, as well as clinical validation and integration is needed for a widespread adoption. Thank you so much for listening to our talk and I hope you have a great rest of your day.