
Building a Smart Campus Operations Assistant with AWS Textract, Bedrock, and Serverless Pipelines
An in-depth technical deep dive into how our team built an AI-powered campus event management platform during the First Commit Hackathon using AWS Textract, Amazon Bedrock, AWS Lambda, and S3.
Introduction & Background
Managing large-scale campus events, departmental orientation programs, and student section rosters manually creates major operational headaches for Class Representatives, event leads, and volunteer coordinators[cite: 3]. Traditional workflows rely heavily on paper sign-in sheets or fragmented CSV spreadsheets[cite: 3]. These methods frequently cause data loss, search latency, and communication breakdowns during busy events[cite: 3].
During the First Commit Hackathon (Bharat Builds Tour), our team engineered the Smart Campus & Event Operations Assistant[cite: 1, 3]. Our objective was to modernize campus event management by combining computer vision document parsing, natural language AI querying, serverless microservices, and real-time task tracking into a unified web application[cite: 3].
Key Operational Challenges Addressed
- Manual Ingestion Bottlenecks: Converting handwritten or printed physical sign-in sheets into digital database records takes hours of manual transcription[cite: 3].
- High-Latency Student Lookups: Locating individual roll numbers or checking student presence across large sections under time pressure is slow and error-prone[cite: 3].
- Rigid Spreadsheet Filtering: Answering basic administrative questions—such as "How many students are missing from Section C?"—requires complex spreadsheet filters[cite: 3].
- Unsynchronized Volunteer Coordination: Event tasks like registration desk duty, entrance crowd control, and certificate distribution are tracked informally, leading to poor operational visibility[cite: 3].
System Architecture & Workflow
The platform utilizes an event-driven, fully serverless architecture hosted on AWS cloud services paired with a modern React/Vite dashboard[cite: 3].
Step-by-Step Implementation & AWS Cloud Services
1. Secure Object Storage with Amazon S3
All uploaded student rosters, CSV datasets, and scanned physical attendance sheet images are ingested directly into private Amazon S3 buckets[cite: 3]. We configured bucket lifecycle rules and access controls to ensure secure handling of sensitive student roll numbers and section lists while maintaining low latency for downstream processing[cite: 3].
2. Automated OCR Pipeline via AWS Textract
To eliminate manual data entry from physical sign-in logs, we leveraged AWS Textract[cite: 3]. When a coordinator uploads a photo or PDF of a printed roster, Textract uses computer vision models to parse tabular structures, student names, and roll numbers[cite: 3].
- The service extracts raw bounding boxes and table key-value pairs[cite: 3].
- A Lambda function converts this output into clean, structured JSON format ready for instant querying on the dashboard[cite: 3].
3. Conversational Intelligence with Amazon Bedrock
Instead of forcing event leads to build database filters manually, we integrated Amazon Bedrock[cite: 3]. The active roster state is passed as a structured context payload to foundational LLM endpoints[cite: 3]. Coordinators can type plain-English questions directly into the AI chat interface:
- "Show all absent students in Section C."
- "What is the current attendance percentage across all sections?"
- "List roll numbers checked in after 10:00 AM."
Amazon Bedrock analyzes the JSON state and returns immediate, accurate, and human-readable operational summaries[cite: 3].
4. Serverless API Orchestration via AWS Lambda & API Gateway
Backend execution is led by AWS Lambda microservices exposed via Amazon API Gateway[cite: 3]. The functions manage file ingestion triggers, parse JSON payload streams from Textract, interface with Bedrock endpoints, and synchronize task status updates across the volunteer task board[cite: 3]. Operating completely serverless ensures zero idle maintenance costs and effortless scaling during peak event check-ins[cite: 3].
Team Engineering Roles & Deliverables
Building this application within the hackathon timeframe required clear task division across full-stack design and cloud infrastructure[cite: 1, 3]:
- Akshansh Pundir (Team Leader): Spearheaded system design, product strategy, and UI/UX engineering[cite: 3]. Built the interactive React frontend with Tailwind CSS, custom liquid glassmorphic components, live stat counters, and background HLS video rendering. Developed the client-side data layer (instant search, section filtering, attendance toggles, and volunteer Kanban board) and connected the Amazon Bedrock conversational query pipeline[cite: 3].
- Anirudh Chauhan (Backend & Data Engineer): Led backend architecture and database pipelines[cite: 3]. Developed serverless API routes using AWS Lambda and Amazon API Gateway, engineered the AWS Textract document parsing workflow, and set up Amazon S3 storage bucket policies and data normalization functions[cite: 3].
Key Technical Results & Metrics
- Extraction Speed: Converted scanned multi-page roster sheets into searchable JSON tables in under 3 seconds using AWS Textract[cite: 3].
- Zero-Latency Search: Enabled instantaneous filtering across thousands of student roll numbers and names[cite: 3].
- Streamlined Operations: Reduced administrative lookup times for event leads by over 80% via Amazon Bedrock natural language queries[cite: 3].
Key Takeaways & Future Expansion
Leveraging AWS managed services allowed our team to move rapidly from design concept to a fully functional, production-ready prototype during the hackathon[cite: 1, 3]. Tools like AWS Textract and Amazon Bedrock abstracted complex machine learning overhead, giving us the freedom to focus on building an intuitive user experience[cite: 3].
Future enhancements planned for the platform include adding real-time WebSocket notifications for volunteer task assignments, automated email alerts for absent students, and multi-region S3 replication for university-wide multi-campus events.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article