How Popsa used Amazon Nova to inspire customers with personalised title suggestions

2026-04-27 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, medium

Summary

Popsa, a technology company operating in over 50 countries, enhanced its Photo Book title suggestion feature by integrating Amazon Bedrock and the Amazon Nova family of models. Initially, Popsa used a graph-based algorithm, "Title Suggestion Graph," since 2021, which relied on photo metadata and on-device CNNs to generate basic titles. In June 2024, Popsa transitioned to a generative AI approach, utilizing Retrieval-based few-shot prompting with Anthropic's Claude 3 Haiku via Amazon Bedrock, resulting in a 13% increase in positive user feedback. Further optimization in early 2025 involved A/B testing with Amazon Nova Lite and Pro, where Nova Pro achieved 73% positive feedback and Nova Lite offered comparable quality to Claude 3 Haiku at significantly lower cost and faster response times. Popsa also reduced "Time to First Suggestion" by 35%, from 1.41 seconds to 0.92 seconds, by migrating to Amazon Bedrock's ConverseStream API, which streams tokens in real time. This system generated over 5.5 million personalized titles in 2025.

Key takeaway

For AI Product Managers evaluating generative AI solutions for content creation, consider Amazon Nova models via Amazon Bedrock. Your team can achieve higher user satisfaction and reduce operational costs and latency compared to other LLMs like Claude 3 Haiku, especially when leveraging the ConverseStream API for improved responsiveness. This approach allows for rapid iteration and measurable uplifts in engagement and purchase rates.

Key insights

Generative AI, specifically Amazon Nova models, significantly improves personalized content generation quality, speed, and cost efficiency.

Principles

• Technology should simplify user experience.
• Iterative testing with LLM-as-a-judge accelerates model selection.
• Streaming APIs enhance perceived responsiveness.

Method

Combine metadata, computer vision, and retrieval-augmented generative AI with few-shot prompting to generate creative, brand-aligned, multilingual titles. Evaluate using strict rules and LLM-as-a-judge for qualitative guidelines.

In practice

• Use Amazon Bedrock for unified API access to multiple LLMs.
• Implement ConverseStream API for faster initial responses.
• Fine-tune smaller models with data from larger models.

Topics

• Generative AI
• Amazon Bedrock
• Amazon Nova Models
• Claude 3 Haiku
• Retrieval-Augmented Generation

Best for: AI Engineer, Machine Learning Engineer, AI Product Manager

Related on AIssential

• How Hapag-Lloyd uses Amazon Bedrock to transform customer feedback into actionable insights — Automating customer feedback analysis with generative AI transforms manual processes into strategic insights.
• Amazon adds AI-generated product previews to search results — Amazon is integrating AI-generated visual previews to enhance product discovery and address imprecise user search queries.
• Cost-efficient custom text-to-SQL using Amazon Nova Micro and Amazon Bedrock on-demand inference — Fine-tuning Amazon Nova Micro with LoRA on Amazon Bedrock provides cost-efficient, production-ready text-to-SQL for custom dialects.
• Transform Enterprise Search with GenAI: Build a Production-Ready RAG Knowledge Assistant Using… — RAG with Amazon Bedrock enhances enterprise search by providing accurate, grounded, natural language access to proprietary data.
• How Amazon Uses LLMs to Recommend Products — Amazon's COSMO system uses LLMs and a multi-stage pipeline to build and serve commonsense knowledge for e-commerce.
• Amazon’s AI Shopping Overhaul — Generative AI is transforming Amazon's shopping experience, enhancing product discovery and personalization.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.