I analyzed AI's understanding of mortgage documents, focusing on machine learning, deep learning, and NLP techniques. I explored how AI processes text and utilized computer vision and OCR to extract information from scanned documents, gaining insights into the mortgage industry's challenges and

I utilized Python to clean and organize mortgage documents, preparing them for AI automation. This involved employing data preparation techniques to standardize information and enhance OCR accuracy through image processing, ensuring that the data was structured and ready for effective AI model use.

I converted both digital and scanned PDFs into machine-readable formats using Python, leveraging tools like PyMuPDF and pdfplumber. I employed field heuristics, including regex patterns and layout clues, to accurately extract key information such as borrower names and loan amounts from mortgage

I conducted a comparative analysis of three OCR engines—Tesseract, PaddleOCR, and EasyOCR—on complex mortgage documents. I learned to extract and clean text from scanned PDFs, evaluated each tool's performance, and made recommendations for optimal OCR solutions based on technical and business needs.

I built a Retrieval-Augmented Generation (RAG) pipeline using LlamaIndex, enhancing AI's ability to retrieve relevant data from large document sets. This project underscored the importance of efficiently processing vast text data and improving the retrieval process within document intelligence.

I advanced my RAG pipeline by implementing precision improvements and experimenting with open-source LLMs. I explored chunking techniques and metadata-based filtering, optimizing the system for better document segmentation and evaluating various models for retrieval accuracy and efficiency.

I developed a system to process and classify unstructured mortgage blobs, separating them into individual documents. I applied layout-based clues and machine learning methods to identify document types, ensuring that each was routed to the appropriate extraction logic for efficient processing.

I created a chatbot integrating my RAG pipeline using Gradio, allowing users to interact with the system via a user-friendly web interface. I modularized my code to facilitate the integration of different models, resulting in a functional chatbot capable of answering queries about PDF documents.

I built my first AI agent, a Moodboard Generator, that transformed prompts into visual outputs. By configuring my workspace in n8n and executing workflows, I grasped how to quickly convert abstract ideas into tangible visuals, laying the groundwork for more advanced agents.

I developed a Consumer Trend Discovery AI Agent that mimicked market analyst behavior. By scanning data from sources like Amazon and design blogs, the agent identified emerging product trends, enabling Wayfair to anticipate market shifts more swiftly than traditional research methods.

I created a Competitor Monitoring AI Agent that analyzed competitor listings, product launches, pricing, and customer reviews. This agent provided insights into competitor strategies and identified opportunities for Wayfair, helping inform data-driven merchandising and marketing decisions.

I enhanced an existing AI Insights & Content Agent into a marketing intelligence system that generated actionable content ideas for Wayfair's marketing team. By refining its logic and tone, the agent produced on-brand campaign ideas rooted in real market insights.

I explored the mechanisms of Machine Learning, Deep Learning, and Natural Language Processing (NLP) to understand how AI interprets pharmaceutical documents. This involved analyzing a real pharmaceutical documentation file to assess AI models' capabilities in managing both structured and

I utilized Python for data processing to prepare pharmaceutical documents for AI automation. This included cleaning and standardizing data, as well as enhancing OCR accuracy through image processing techniques, ensuring the data was structured effectively for subsequent AI model applications.

I converted digital PDFs into machine-readable formats using Python tools like PyMuPDF and pdfplumber. By applying field heuristics, I identified key document fields such as manufacturing dates and vendor names, laying the groundwork for compliance workflows to flag outdated documents.

I conducted a comparative analysis of three OCR engines—Tesseract, PaddleOCR, and EasyOCR—on scanned pharmaceutical documents. I assessed text extraction methods and layout-aware tools to recommend the most effective OCR solutions for real-world automation scenarios.

I built a Retrieval-Augmented Generation (RAG) pipeline using LlamaIndex, enabling efficient retrieval of relevant information from extensive document sets. This project improved the AI's ability to process large volumes of text while ensuring accuracy.

I enhanced the RAG pipeline by implementing advanced chunking techniques and metadata-based filtering. This refinement improved document segmentation and retrieval accuracy, allowing for the evaluation of open-source models against proprietary options.

I designed a system for processing large, unstructured pharmaceutical regulatory files, splitting them into individual documents and classifying them for the appropriate extraction workflows. This modular approach aligns with real-world pharmaceutical document management.

I developed an interactive chatbot using Gradio, integrating it with my local RAG pipeline. This feature allowed users to query the system through a user-friendly web interface, enhancing accessibility and interaction with the document intelligence system.

I examined case studies of FedMart, Cadbury, and WeWork, mastering the Incorruptible 4-Lens Framework. I selected a company that failed and created an Incorruptible Scorecard, breaking down the factors that led to its downfall, and produced a one-minute video summarizing my findings.

I reflected on the elements of the Incorruptible framework that resonated with me personally, completed the Life Pie exercise to define my fulfillment, and selected a company I admire. I evaluated this company through the 4 lenses and created my own Incorruptible Scorecard, preparing a thoughtful

I analyzed the IoT systems of The Grand Marina Hotel, mastering the CIA Triad and identifying six major attack types. By the end of the week, I delivered a comprehensive threat model, laying the groundwork for future security measures.

I learned Python fundamentals and applied them to IoT security data, exploring Google Colab and mastering data preparation techniques with JSON and Pandas. This culminated in writing scripts to analyze HYDROLOGIC sensor data, equipping me with essential programming skills for the project.

I constructed a complete IoT pipeline, including data publishing from sensors and real-time message routing. By intercepting my own traffic, I gained firsthand experience with the dangers of unsecured MQTT communications.

I enhanced the MQTT pipeline with TLS encryption, securing the transmission of water sensor data. Through experiments, I evaluated the security-performance tradeoff, ultimately providing a professional security assessment for the hotel's General Manager.

I implemented mutual TLS (mTLS) to enforce device identity within the secured pipeline. This process ensured that only authorized devices could connect to the broker, resulting in the creation of a provisioning policy document for The Grand Marina.

I developed three layers of defense against replay attacks, including timestamp validation and HMAC message signing. This work ensured that the subscriber could accurately verify the freshness and authenticity of each message received.