EDU Assist

Find the best college for you using this conversational interface.

Comment

💡 Inspiration

My kids will be searching for a college in a few years, so I want to provide them with a conversational resource grounded with reliable data. Finding the right college is a crucial decision that can shape a student's future, and having access to trustworthy, comprehensive information in an easy-to-use format is essential.

🎯 What EDU Assist Does

EDU Assist provides a conversational interface to help students find the right college for them. It is grounded via tools that wrap the US Department of Education College Scorecard API, ensuring that all information comes from official, reliable sources.

Key Features:

Conversational Search: Ask questions in natural language about colleges and programs Reliable Data: All information comes from the official U.S. College Scorecard database

Comprehensive Coverage: Search by location, program type, admission rates, costs, and outcomes Real-time Responses: Get instant answers with streaming chat interface

Educational Focus: Designed specifically for college search and planning

🛠️ How It's Built

I used Strands Agents to build a Python agent and set of tools that integrate with the Scorecard API, powered by Claude Sonnet via Amazon Bedrock. The system combines modern AI capabilities with trusted educational data sources.

Technologies Used: 🤖 Strands Agents 🐍 Python 🧠 Claude Sonnet (Anthropic) ☁️ Amazon Bedrock ⚡ FastAPI 🌐 HTML/CSS/JavaScript 📊 U.S. College Scorecard API 🔄 Server-Sent Events (SSE) 📡 RESTful APIs 🐳 Docker 🚀 AWS App Runner

🚧 Challenges Overcome

Strands kept outputting raw tool JSON data in the response. I built another agent with tools hooked into a different API and it didn't repeat. After reading all the docs and much debugging, I simply told Strands to not output all the Tool JSON in the system prompt and that worked!

🏆 Accomplishments

This was my first Strands agent (other than a demo build). It uses tools to leverage an existing API, the College Scorecard, to create a conversational experience that makes complex educational data accessible through natural language.

📚 What I Learned

Amazon Bedrock: How to use Amazon Bedrock as a model provider for AI applications Strands Agents: How to build an agent with Strands Agents framework API Integration: How to ground AI responses using tools that wrap existing APIs Trusted Data Sources: Working with reliable, government-provided educational datasets Conversational AI: Creating natural language interfaces for complex data queries

🔮 What's Next

The US Bureau of Labor and Statistics has an API that provides wage data. This would allow users to understand how well a particular career might pay. Although the general LLM does a pretty good job translating a user's description of a degree program to an actual degree program, I might be able to improve results by embedding a full list of degree programs.

Future Enhancements:

Career Outcomes: Integration with Bureau of Labor Statistics for salary data Program Embeddings: Enhanced degree program matching through embeddings Personalized Recommendations: AI-driven college matching based on preferences Financial Planning: Cost analysis and financial aid guidance

Built With

Share this project:

Updates