GenAI-Powered Review Summarization and Visualization
A serverless AWS pipeline that uses Amazon Bedrock model to transform raw customer reviews into structured insights. It automates summarization, sentiment tagging, and dashboard visualization, enabling faster decision-making with minimal operational overhead.
Overview
2025,
Usecase
This project demonstrates how e-commerce businesses can convert unstructured customer feedback into actionable insights using AWS Generative AI. The pipeline ingests CSV review data from S3, processes it through Lambda and Bedrock for structured summarization, and delivers outcomes into QuickSight dashboards. By combining serverless orchestration with AI-powered text analysis, the system reduces manual review effort while improving visibility into customer sentiment and product performance.
Process
1. S3 Setup – Reviews uploaded to input_reviews/, batches handled via Lambda.
2. Batching Lambda – Splits large files into 25-review chunks stored in review_batches/.
3. Summarization Lambda – Calls Bedrock Titan with structured prompts and stores results in summarized_batches/.
4. Auto-Merge Lambda (EventBridge) – Hourly merge of summaries into cleaned_reviews.json.
5. Flattening Lambda – Generates cleaned_reviews_flat.csv and feature_sentiment_exploded.csv for dashboards
6. Visualization – QuickSight dashboards show feature sentiment, pros/cons, and structured review tables
Outcome
* Automated the full review analysis pipeline from ingestion to visualization.
* Produced structured outputs (summaries, pros, cons, feature sentiment) ready for business reporting.
* Enabled QuickSight visuals like bar charts, word clouds, and summary tables.
* Reduced manual review effort to minutes, delivering faster insights for analysts and product teams.
* Showcased a scalable, cost-efficient GenAI solution leveraging AWS-native services.
Nikhil Varma
ML Engineer
2025,
Case-Study
