Agentic Scent Discovery Engine
The Problem
Luxury perfume buyers struggle with 'choice paralysis' online. Standard filters (e.g., 'floral') are too generic, leading to abandoned carts and low conversion for high-intent customers.
The Solution
Built a conversational AI agent that acts as a personal fragrance consultant. It asks nuanced questions about mood, desired memories, and personality traits, then translates those abstract concepts into concrete scent profiles using a multi-step LLM chain.
The Impact
Reduced discovery time by 80%, increased 'add to cart' events by 35% in user testing, and provided a highly differentiated, luxury brand experience.
Automated E-commerce Data Pipeline
The Problem
A growing e-commerce store was manually categorizing thousands of new products monthly. This was slow, error-prone, and required significant human hours, creating a bottleneck for scaling their inventory.
The Solution
Developed an automated data pipeline using Google Cloud Functions and a multimodal LLM. The system ingests product images and descriptions, auto-generates SEO-optimized titles, assigns them to categories with 98% accuracy, and flags low-quality images for review.
The Impact
Eliminated 120 hours of manual data entry per month, accelerated time-to-market for new products by 95%, and improved SEO ranking through consistent, high-quality metadata.