In this project, you’ll build your first AI agent: a Moodboard Generator that turns simple prompts like “modern living room with warm colors” into visuals. By setting up your workspace in n8n and running your first workflow, you’ll see how ideas can instantly become outputs. What feels like a creative experiment is actually the foundation of Wayfair’s edge: turning abstract trends into concrete visuals fast. This hands-on win sets you up to build the more advanced agents that will track market shifts, monitor competitors, and generate insights that matter for Wayfair..

I built a Consumer Trend Discovery AI Agent that emulates market analysts by scanning real-world data to identify rising product trends. This project aimed to enhance Wayfair's ability to detect emerging design preferences swiftly, informing smarter merchandising decisions.

In this project, you’ll shift from trend discovery to competitive intelligence: teaching your agent to think like a market strategist. Your agent will collect and analyze competitor product listings, new launches, price movements and reviews. The goal: to uncover what competitors are doing differently , how their customers are responding, and where Wayfair’s whitespace opportunities might lie. By the end of this project, your Competitor Monitoring AI Agent will surface actionable insights helping teams benchmark Wayfair’s products, spot early shifts in market positioning, and make faster, data-backed merchandising and marketing decisions.

I enhanced a pre-existing AI Insights & Content Agent by studying Wayfair's brand voice and refining its output. I customized the agent to generate on-brand marketing content, selecting new content types and improving the tone, making it a valuable tool for the marketing team.

I summarized how large language models process text, covering tokenization, embeddings, transformers, and training dynamics.

I prepared and cleaned pharmaceutical-style documents using Python, applying data cleaning, JSON handling, text normalization, and image preprocessing across Colab notebooks.

I built a Python pipeline to extract and structure text and field values from multi-page pharmaceutical SDFs using PyMuPDF and positional heuristics.

I compared Tesseract, PaddleOCR, and EasyOCR on a scanned pharmaceutical certificate and documented differences in detection, confidence, and setup effort.

I built and optimized a Retrieval-Augmented Generation pipeline using LlamaIndex, implementing chunking, embeddings, hybrid retrieval, and reranking for precise document QA.

I tested three RAG configurations and compared their retrieval behavior on a sample SDF document.

I built a page-level segmentation and metadata extractor for bundled PDF files, producing labeled page outputs and doc boundary flags.

I built a Gradio chatbot interface and connected it to a local RAG pipeline for document Q&A, with UI controls and source attribution.