LynxAI

Inspiration Most planners just list deadlines; we wanted an AI "Chief of Staff" that understands the math of a GPA, helping students overcome decision fatigue by identifying which tasks actually move the needle. What it does LynxAIuses Agentic AI to perform a real-time cost-benefit analysis of a student's time, syncing with Canvas to recommend the highest-ROI tasks whenever a gap opens up in their schedule. How we built it We used LangGraph to orchestrate a triple-agent system (Harvester, Analyst, Strategist) powered by Gemini 1.5 Flash, with a FastAPI backend and a RAG pipeline to parse grading policies from syllabi. Challenges we ran into Normalizing unstructured syllabus data—like varying late policies and weighted categories—into a consistent logic for our priority algorithm required intensive prompt engineering and RAG tuning. Accomplishments that we're proud of The "Chaos Button": a demo feature that simulates a cancelled class, triggering our agents to autonomously re-evaluate the entire day and serve up a new "Current Mission" in seconds. What we learned We discovered the power of Agentic Workflows over single prompts; decoupling data retrieval from reasoning allowed for a much more reliable and "math-aware" AI assistant. What's next for LynxAI Refining the Ghost Writer feature to proactively draft professor emails and integrating "Energy Level" tracking to balance high-intensity studying with burnout prevention.

Built With

• gemini
• json
• python
• v0