SurveySynth: Intelligent Feedback Analyzer

SurveySynth – Transforming Student Feedback into Actionable Insights, Instantly

Comment

SurveySynth – Serverless Survey Analyzer for Education

By: Vishnu Murthy | B.Tech Final Year | KL University
Deployed Frontend: SurveySynth UI (S3)

Problem Statement

In large universities like mine (KL University), thousands of Computer Science students share the same set of core courses. Each course is managed by a course coordinator who gathers post-course feedback to improve the syllabus and delivery for future batches.

However, manually analyzing CSV-based survey responses (often hundreds or thousands of entries) is inefficient and subjective. There’s no automated system to extract pain points, sentiment, or actionable suggestions from open-text feedback.

Solution: SurveySynth

SurveySynth automates this end-to-end process using serverless architecture, enabling coordinators to:

  • Upload raw CSV survey files
  • Automatically clean and analyze data
  • Generate AI-powered summaries and suggestions
  • View visual charts and download reports

Key Features

  • User Registration & Login (React + API Gateway + Lambda + DynamoDB)
  • CSV Upload → Stored in S3 → Triggers ETL
  • AWS Glue preprocesses the data and stores it in a refined format
  • Amazon Bedrock analyzes textual feedback for insights
  • AWS Glue generates Seaborn/Matplotlib charts for dashboard
  • Exportable PDF reports

Architecture Overview

Architecture Diagram

Built 100% using Serverless Technologies on AWS

1. Register/Login → API Gateway → Lambda
2. Upload CSV → S3 → uploadSurveyHandler → DynamoDB
3. Triggers Glue ETL Job → Cleans & Writes Silver CSV
4. Lambda updateStatusPostGlue updates metadata
5. summarizeFeedbackHandler uses Bedrock for insights
6. Results saved to SurveyInsights table
7. Stream triggers triggerVisualizationJob → Glue → PNG charts
8. Frontend polls for results → renders summary + visuals

🔗 Endpoints & Lambda Flows

Flow Step Trigger/Service Lambda Function Purpose
Register/Login API Gateway UserApiHandler Auth (Mock, stores in DynamoDB)
Upload Survey API Gateway uploadSurveyHandler Upload to S3, update SurveyMeta
Preprocess CSV S3 → Glue Trigger SurveyPreprocessJob Clean CSV → silver/ folder
Update ETL Status EventBridge updateStatusPostGlue Set status = preprocessed
Generate AI Summary Lambda Stream summarizeFeedbackHandler Bedrock → SurveyInsights
Generate Charts DynamoDB Stream triggerVisualizationJob Triggers Visual_job Glue Job
Final PNG Charts Glue Job Visual_job Generate & upload Seaborn/Matplotlib
📡 Fetch Data for UI API Gateway SurveyMetaHandler, SurveyInsightHandler, getChartUrlsHandler Feed UI

Highlights

Language Support: Works with free-text fields in English

Scalability: Serverless — scales with number of students/surveys

AI Integration: Amazon Bedrock → Claude Model for summarization

Visualization: Seaborn-based PNG charts (instead of QuickSight)

Deployment: Fully deployed via AWS SAM and S3

AWS Services Used

Service Usage
Lambda Business logic: 8+ functions (auth, upload, summarize, chart gen)
API Gateway Login, Register, Upload, Insights APIs
DynamoDB Users, SurveyMeta, SurveyInsights metadata
S3 Stores raw/silver CSVs, PNG charts, optional PDF
AWS Glue 2 jobs: SurveyPreprocessJob (cleaning) and Visual_job (charts)
Amazon Bedrock Summarization and sentiment analysis using Claude model
EventBridge Detect Glue job completion
CloudWatch Logging, debugging during development
CloudFormation Serverless Application model

Frontend Flow (React + S3 Hosted)

  1. Register/Login → Save session in localStorage
  2. Upload CSV file (base64) → /upload endpoint
  3. Polls /survey-meta and /insights APIs
  4. Displays:
    • Summary
    • Actionable Suggestions
    • PNG Charts (fetched from S3)
  5. Frontend is deployed at: 👉 http://buck30300.s3-website-us-east-1.amazonaws.com/ You can Use These Credentials instead of signup: (srivis2005@gmail.com && 123456) || (srivis2001@gmail.com && 123456)
  6. https://github.com/murthy30300/SurveySynth/blob/feature/master/synthetic_students_feedback_500.csv
  7. use this csv file for upload, it will take less than 3 minutes of time at least.
  8. Note: the above csv file generated by GPT, the comments may not be accurate in data but according the sentiment of data in csv file the charts can be generated according the sentiment.

Example Output (from 10 feedback rows)

{
  "summary": "Users found the interface intuitive but faced performance issues.",
  "pain_points": ["App crashes", "Slow load times", "Login bugs"],
  "positive_feedback": ["Great design", "Useful search function"],
  "actionable_suggestions": [
    "Fix login crashes",
    "Improve app speed",
    "Add dark mode"
  ]
}

Local Setup (if required)

  1. Setup S3
  2. create Dynamodb table
  3. npm install # React frontend
  4. npm run dev # Runs at http://localhost:5173
  5. sam build && sam deploy --guided # Deploy backend

Future Enhancements

  1. Generate downloadable PDF Reports
  2. Embed charts inside PDF
  3. Add email notification on upload completion
  4. Switch to Amazon Cognito for full auth
  5. Multi-language summarization

About the Developer

Vishnu Nukala University ID: 2200030300 Final-year B.Tech | KL University Used SurveySynth to solve a real problem at scale — improving courses for thousands of students based on feedback that matters.

Built With

Share this project:

Updates