Verdict – Know Before You Buy

Online fashion shopping has become incredibly convenient, but it's still difficult to answer one simple question before buying: Will this actually suit me? Most shopping platforms focus on products, brands, and trends, but very few help users make confident purchasing decisions based on their own appearance, style, and the occasion. We wanted to build an AI-powered decision support system that helps people understand why an outfit works or doesn't work before spending money. Instead of creating another chatbot or virtual try-on application, we focused on building an explainable AI product where every recommendation is transparent and supported by reasoning. Verdict – Know Before You Buy is an AI-powered fashion purchase decision assistant. Users upload: Their current photo A garment photo The occasion Optional style preferences Verdict analyzes the outfit across six evaluation dimensions: Occasion Fit Color Harmony Formality Seasonality Style Consistency Style Preference Match The application combines AI-generated analysis with deterministic application logic to produce an explainable verdict. It also generates a privacy-first Style DNA profile that captures long-term styling characteristics without storing personal photos. Finally, the AI Shopping Advisor recommends better clothing alternatives, explains why each recommendation fits the user, and provides retailer search links for further exploration. We built Verdict using a modern full-stack architecture focused on explainability, privacy, and maintainability. Frontend Next.js React TypeScript Tailwind CSS AI OpenAI Responses API GPT-4.1 Vision Structured JSON outputs Runtime schema validation Application Logic Deterministic verdict calculation Explainable decision pipeline Privacy-first Style DNA AI Shopping Advisor Retail search link generation Throughout development, Codex acted as an engineering collaborator. It helped us: Plan the project architecture Extend runtime schemas Implement frontend and backend features Generate and refine tests Review implementation milestones Debug issues while preserving the existing architecture The biggest challenge was balancing AI flexibility with predictable application behavior. We wanted AI to provide rich reasoning while ensuring the final verdict remained deterministic and reproducible. Other challenges included: Designing structured AI outputs Runtime validation of AI responses Building a privacy-first Style DNA system Keeping shopping recommendations independent of the deterministic verdict Extending features without breaking previous phases We're proud that Verdict evolved beyond a simple AI demo into a complete product experience. Highlights include: Explainable AI decision making Deterministic verdict engine Privacy-first Style DNA AI Shopping Advisor Runtime validation for AI outputs Production-ready architecture Live deployment Comprehensive testing across the analysis pipeline This project reinforced that building reliable AI products requires much more than calling an LLM. We learned the importance of: Combining AI reasoning with deterministic application logic Validating AI outputs before using them Designing for explainability rather than automation alone Building privacy-first AI experiences Using Codex as an engineering partner throughout iterative development We see Verdict evolving into a complete AI-powered fashion decision platform. Future plans include: Wardrobe Memory for previously analyzed clothing Personalized capsule wardrobe recommendations Outfit planning across multiple occasions Seasonal wardrobe insights Multi-item outfit comparison Continuous Style DNA refinement Smarter recommendation diversity Expanded explainability and personalization Our long-term vision is to help people make more confident, informed clothing purchase decisions while keeping AI transparent, trustworthy, and privacy-first.