GrabMaps’ Post

📌 I’m pleased to share the completion of the Associate AI Engineer for Data Scientist Track, a transformative program that deepened my expertise in AI and data science. The coursework covered advanced topics such as machine learning (ML), MLOps, explainable AI, deep learning, and large language models (LLMs), alongside essential skills in software engineering principles and Python testing. Through hands-on projects, I gained practical experience in building and deploying scalable, efficient AI systems. This achievement has significantly enhanced my technical expertise and strategic understanding of AI-driven systems. I’m excited to leverage these skills to drive innovation, solve complex problems, and contribute to impactful AI initiatives. #ArtificialIntelligence #MachineLearning #DeepLearning #MLOps #LLMs #DataScience #ProfessionalDevelopment #AIInnovation

Excited to start the new year on a positive note! 🚀 Sharing my initial milestone on the AI Engineer career path with #DataCamp. 🧠📊 Looking forward to learning, growing, and diving deeper into the world of AI in 2025. 💡✨ #CareerGrowth #DataScience #LearningNeverStops #ArtificialIntelligence

Be Aware you are at Risk⚡ I was just going through my feeds and saw much news about the deep fake again. We need to control it; it's now again out of control, and this and that. So, after going through some sources, I built a basic Gradio app and deployed it on hugging faces. Deepfake Detection App! 🎉 With the rise of deepfake technology, identifying manipulated images has become increasingly challenging. To combat this, I developed a web application that leverages state-of-the-art AI to detect deepfakes with high accuracy. 🔍 How It Works: Face Detection: Utilizes the MTCNN model to detect faces in an image accurately. Face Classification: Employs a pre-trained InceptionResnetV1 model to classify faces as real or fake. Explainability: Uses Grad-CAM to visually explain the model's decision, ensuring transparency and trust in the predictions. 🌟 Key Features: Easy-to-use interface powered by Gradio. Real-time deep fake detection with confidence scores (needs to be worked on) Visual explainability to understand model decisions. 📈 Try It Out: The app is now live on Hugging Face Spaces! Check it out, upload an image, and see the magic happen. 🔗 Deepfake Detection App on Hugging Face https://lnkd.in/dys6RKnE 💡 Technical Details: Built with PyTorch, FaceNet-PyTorch, and PyTorch-Grad-CAM, this app showcases the power of AI in tackling modern-day challenges. Github code: https://lnkd.in/dzVmEbC8 🙏 Acknowledgments: A big thank you to the open-source community for providing the incredible tools and libraries that made this project possible. 🚀 Ready to dive into the world of AI and deepfake detection? Try out this app and share your feedback! #AI #MachineLearning #DeepLearning #DeepfakeDetection #Gradio #PyTorch #HuggingFace #ArtificialIntelligence #Innovation

Day 25 of AI 🚀 From Kaggle Competitions to Real-World Solutions: Today’s Wins! 📊🩺 Thrilled to share two exciting milestones in my machine learning journey: 1️⃣ Insurance Prediction on Kaggle Joined the "Insurance Prediction Challenge" on Kaggle, diving into actuarial data to forecast insurance costs. Tackled feature engineering, hyperparameter tuning, and model stacking (shoutout to XGBoost and RandomForest) - 2️⃣ Deploying Medical AI: Pneumonia Detection Web App 🔍🌐 Turned my CNN model for chest X-ray analysis into a user-friendly web app! Now, anyone can upload an X-ray image and instantly see if pneumonia is detected. Here’s how it works: - Tech Stack: Flask backend + TensorFlow model - Impact: Early detection saves lives, and democratizing access to AI-driven diagnostics feels incredibly rewarding. - Will do Inshallah: Grad-CAM visualization (to *show* where the model "looks" in the X-ray). Why This Matters - Bridging gaps: From Kaggle notebooks to deployed apps, the end-to-end ML workflow is now clearer than ever. - AI for good: Proud to contribute to tools that empower healthcare decisions, even in small ways. Next Steps - Improve the app’s accessibility (mobile-friendly UI, multilingual support). - Explore federated learning to enhance the model across diverse datasets without compromising privacy. Huge thanks to the open-source community for tools like TensorFlow, Flask, and Grad-CAM! #AI #DataScience #HealthcareAI #MachineLearning #WebDevelopment #Kaggle Curious about the code or want to collaborate? Let’s connect!👇

Excited to share the journey behind our latest project, MedBot, developed in collaboration with Abdul Munim Baig. We’ve created two versions of MedBot—one powered by the 𝗨𝗻𝘀𝗹𝗼𝘁𝗵 Meta 𝗟𝗹𝗮𝗺𝗮 𝟯.𝟭 𝟴𝗕 𝗾𝘂𝗮𝗻𝘁𝗶𝘇𝗲𝗱 𝗺𝗼𝗱𝗲𝗹 for fast inference, and another using the Google 𝗚𝗲𝗺𝗶𝗻𝗶 𝟭.𝟱 𝗙𝗹𝗮𝘀𝗵 𝗔𝗣𝗜. 🔍 𝗢𝗯𝗷𝗲𝗰𝘁𝗶𝘃𝗲: Our goal was to create reliable medical assistants using advanced AI technology. Both versions of MedBot are designed to address a wide range of medical queries, from everyday health questions to more urgent concerns. 📚 𝗗𝗮𝘁𝗮 𝗮𝗻𝗱 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴: For training, we utilized the Malikeh1375/medical-question-answering-datasets from Hugging Face, providing a robust foundation for fine-tuning our models and ensuring that MedBot delivers accurate information. 💻 𝗖𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀: Fine-tuning large models like Llama 3.1 8B presented challenges, particularly with GPU RAM requirements. By implementing 4-bit quantization, we achieved fast inference speeds, making it feasible to deploy alongside the Gemini 1.5 Flash API model. 🚀 𝗨𝘁𝗶𝗹𝗶𝘇𝗶𝗻𝗴 𝗞𝗮𝗴𝗴𝗹𝗲 𝗮𝗻𝗱 𝗗𝗲𝗽𝗹𝗼𝘆𝗺𝗲𝗻𝘁: We used Kaggle’s free GPU resources and ngrok to convert the Kaggle server URL and port into a publicly accessible API for our backend. This approach allowed us to deploy this version of MedBot, ensuring accessibility and efficiency. 🔧 𝗟𝗶𝗺𝗶𝘁𝗮𝘁𝗶𝗼𝗻𝘀 𝗮𝗻𝗱 𝗙𝘂𝘁𝘂𝗿𝗲 𝗪𝗼𝗿𝗸: While the Llama 3.1 8B model excelled in fast inference, it faced some limitations in context maintenance and text streaming. However, both the Llama 3.1 and Gemini 1.5 Flash API versions of MedBot have proven valuable, and we’re eager to continue refining their capabilities. 👨🏫 𝗔𝗰𝗸𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲𝗺𝗲𝗻𝘁𝘀: We are incredibly grateful to Sir Muhammad Farrukh Bashir and Dr. Hammad Majeed for their invaluable guidance throughout the project. Working under their supervision at Genesys Research Lab provided us with a rich learning environment and an opportunity to grow both professionally and personally. 🔧 𝗧𝗲𝗰𝗵 𝗦𝘁𝗮𝗰𝗸: •     𝗙𝗿𝗼𝗻𝘁𝗲𝗻𝗱: HTML, CSS, JavaScript •     𝗕𝗮𝗰𝗸𝗲𝗻𝗱: Python, Flask, TensorFlow, Torch 👨💻 𝗖𝗼𝗻𝗰𝗹𝘂𝘀𝗶𝗼𝗻: The development of MedBot highlights the complexities of AI in healthcare, especially when balancing model performance and computational efficiency. We're proud of the progress we’ve made and look forward to further advancements in this space. #AI #MachineLearning #MedBot #HealthcareAI #TechInnovation #MLOps

🚨 Today’s the day! 🚨 Our free webinar kicks off in just a few hours! Get ready to dive into the future of software engineering in the age of AI. 🕒 Time: 4:00 PM CET 🔗 Join here: https://lnkd.in/daxujZ_a Don't miss out on this opportunity to revolutionize your testing processes. See you there! 🌟 #AI #Webinar #TechRevolution

“What we see more and more from the market is developers and businesses combining small and large models to build the best product experience at the price and the latency that makes sense for them”. Interesting to see small size and mini models gaining speed and making #hybridAI real. #ArtificialIntelligence #MachineLearning #DataScience

🚀 We are thrilled to share our team's contribution to the Kaggle Gemini Competition! With my talented colleague Ramanathan Srinivasan , we've developed a conversational experience for analyzing ClinicalTrail.gov using Google's Gemini AI. 🔍 Key Features: - Implemented a unified caching system for faster query processing - Developed a comprehensive performance analysis framework - Built scalable solutions handling various dataset sizes - Created a detailed visualization system for performance metrics 💡 Technical Highlights: - Utilized Google's Gemini-1.5-Pro-002 model - Implemented intelligent context caching - Built robust error-handling mechanisms - Created dynamic performance visualization dashboard - Long context window (2M) and Context Caching 📊 Results: - Significant query time reduction with caching - Scalable processing of sizeable clinical trial datasets - Detailed performance metrics and error analysis ✨ Special Thanks: - Google Gemini team for providing such a powerful AI model - Kaggle community for the platform and invaluable feedback 🔗 Check out our complete code and documentation on Kaggle: https://lnkd.in/g4e9BTby I am looking forward to hearing your thoughts and suggestions. Let's connect and discuss AI applications in healthcare more! #clinicaltrails #python #google #geminiai #ai #kaggle #gemini #generativeai

Check out our contribution to a kaggle competition. “Clinical insights, delivered at conversation speed” with Angu S K. #GeminiAI #Kaggle #ClinicalTrials #ClinicalData #ConversationalAI #HealthcareInnovation #AIinnovation

Hello Connections🤖 I'm thrilled to introduce my latest project: a cutting-edge Data Science QnA chatbot! 🤖✨ This advanced chatbot leverages the powerful combination of ChatGroq inference and the Llama3 model to deliver accurate and insightful responses to all your data science queries. Experience it here: https://lnkd.in/g4dJnJWn 🌟 Key Features: 🔹 Advanced AI Capabilities: Powered by Llama3, the chatbot understands and responds to data science questions based on Documents provided and trained. 🔹 Efficient Inference: Utilizing ChatGroq inference ensures rapid response times, making interactions seamless and efficient. 🔹 User-Friendly: Designed to be intuitive and easy to use, providing an accessible resource for both beginners and seasoned professionals. Whether you're exploring machine learning, delving into statistical analysis, or simply looking to expand your data science knowledge, this chatbot is here to assist. 📢 I invite you all to try it out and experience the future of data science assistance. Your feedback and insights will be invaluable as we continue to enhance and refine this tool. #DataScience #AI #MachineLearning #Innovation #Chatbot #Llama3 #ChatGroq #TechInnovation #Llama3 #groqapi #DeepLearning #LLM #ml #dl #Huggingface