Inspiration
We were inspired by the struggle of managing time and energy during intense workloads, like hackathons or school. Productivity apps exist, but most don’t adapt to the reality of how people actually feel throughout the day. We wanted to create something that could learn from a person’s schedule and energy levels to build routines that are both effective and sustainable.
What it does
Radnel is a smart scheduling assistant that:
Uses reinforcement-style scoring to rank and prioritize tasks based on energy, urgency, and importance.
Dynamically generates schedules that adjust to how you’re actually feeling.
Uses RAG (Retrieval-Augmented Generation) to pull relevant tasks or events from your data (like emails or calendars) and suggest the best placement in your day.
Provides a clean interface where users can view and edit their schedules in real time.
How we built it
Backend: We used Mastra to build an agent that processes natural language prompts into structured tasks and schedules. Frontend: Built with React, providing an interactive calendar that updates as the agent reschedules tasks.
RAG: We embedded user data into vectors and used dot products to retrieve the most relevant tasks, making recommendations smarter.
Integration: Connected backend and frontend through a REST API for a smooth demo experience.
Challenges we ran into
Debugging endless issues with missing modules and API routes (like the infamous Cannot GET /ping error).
Coordinating tasks across the team — sometimes some members had too much to do while others were idle.
Time pressure: implementing both reinforcement scoring and RAG in 36 hours was ambitious.
Managing energy levels ourselves — it’s hard to build a product about balance when you’re running on 3 hours of sleep.
Accomplishments that we're proud of
Built a working demo that showcases how Radnel creates and updates a schedule.
Successfully integrated Mastra and a React frontend into one project.
Learned how to apply RAG and embeddings in a practical way.
Pushed through challenges as a team, keeping morale high even at 2–3 AM.
What we learned
The importance of clear team communication and distributing work evenly.
How to design and implement a simple reinforcement scoring system for scheduling.
How to use vector embeddings and similarity search to retrieve the most relevant data.
That sometimes, making a clean demo matters more than building every feature perfectly.
What's next for Radnel
Integrating directly with more platforms such as Apple Calender, Canvas and Slack
Using more advanced reinforcement learning to adapt based on user feedback.
Building mobile support so schedules can adapt on-the-go.
Expanding the RAG pipeline to handle larger datasets (documents, to-do lists, etc.).
Log in or sign up for Devpost to join the conversation.