Inspiration
In higher education, academic advisors and college administrators spend up to 70% of their operational hours manually compiling attendance spreadsheets, checking exam marks, and reviewing extracurricular logs across fragmented software portals. Because these processes are manual and reactive, at-risk students are often identified only after failing midterms or breaching critical attendance boundaries. We were inspired to build the Student Growth Agent to serve as a "Good Neighbor" AI that runs silently in the background, handling administrative telemetry tracking autonomously while providing personalized, daily habit guidance to students and escalating to human advisors only when crucial decisions are required.
What it does
It transforms reactive college administration into proactive support. Operating silently in the background, it monitors student telemetry—such as attendance trends, academic performance, and extracurricular commitments. Every day, it formulates personalized, actionable 1–2 sentence micro-tips sent directly to students to help them balance study habits with personal goals. To prevent dashboard fatigue for staff, the agent runs autonomously and only surfaces an automated Human-in-the-Loop decision ping when a critical risk threshold is breached
How we built it
We engineered our solution as a full agent harness using the open-source Strands Agents SDK combined with Amazon Bedrock which uses Nova lite model.For database SQLite and for interface used Streamlit.
Challenges we ran into
Working with Strand's SDK was a completely new experience. Initializing the project and setting up the core architecture presented a unique learning curve. Furthermore, connecting Amazon Bedrock and configuring its model resulted in several initial integration errors. Troubleshooting these roadblocks was challenging, but it ultimately provided a deep, hands-on understanding of the execution pipeline and how the models perform in a complete end-to-end process.
Accomplishments that we're proud of
Successfully built and deployed a fully functional MVP featuring a dual-view Streamlit dashboard that serves both students and advisors. We are incredibly proud of configuring complex multi-agent orchestration using strands-agents-sdk connected to real-time SQLite student telemetry. We also successfully engineered a custom Strands deterministic execution hook for advisor risk pings and released the entire project as an open-source repository under the MIT License.
What we learned
Through building this project, we gained hands-on experience structuring production AI agent harnesses using the Strands Agents SDK. We learned how to write clean, docstring-driven @tool functions, effectively manage multi-agent handoffs, and integrate vital human-in-the-loop workflows for community organizations and educational institutions.
What's next for Student Growth Agent
LMS Integration: Connect directly with Canvas, Blackboard, and university SIS APIs for real-time telemetry streaming. Amazon Bedrock AgentCore Deployment: Further scale background orchestration and compute by deploying the harness to the Amazon Bedrock AgentCore Runtime.
Log in or sign up for Devpost to join the conversation.