FreshCart.

We built FreshCart around three AI systems that work together to optimize the entire customer journey from browse to doorstep: a recommendation engine, a dynamic pricing optimizer, and a delivery logistics system. The recommendation engine uses collaborative filtering combined with contextual signals to predict what each customer wants. It analyzes purchase history, browsing patterns, time of day, day of week, weather, local events, and dietary preferences to generate personalized product suggestions at every touchpoint — the home screen, category pages, search results, and the cart itself. The "Complete Your Meal" feature detects recipe patterns (buying pasta, tomatoes, and garlic suggests the customer is making Italian) and recommends complementary items. The "Running Low" feature predicts when customers are likely to need replenishment based on their historical purchase cadence for consumable items. The dynamic pricing optimizer balances three objectives: maximizing basket size, minimizing cart abandonment, and maintaining healthy margins. It uses reinforcement learning to test and learn optimal pricing strategies across customer segments, adjusting prices for non-essential items, bundling complementary products at slight discounts, and offering personalized promotions triggered by cart composition and abandonment risk signals. Importantly, staple items maintain consistent pricing to preserve trust — the dynamic pricing applies only to discretionary purchases and promotional offers. The delivery logistics system transformed a manual dispatch operation into an automated, real-time optimization engine. When a customer selects a delivery window, the system evaluates current driver locations, pending orders, predicted order volume for the next two hours, traffic patterns, and geographic clustering to assign the order to the optimal driver and route. Routes recalculate continuously as new orders arrive, using a custom vehicle routing algorithm that minimizes total fleet driving time while honoring every customer's delivery window. We integrated a real-time tracking experience so customers see exactly where their delivery is, with AI-generated ETAs that are accurate to within 3 minutes.