CAT: CollaborAiTor for Student Teams

Inspiration

Project-based classes often struggle not because students lack talent, but because teams are formed with limited visibility into schedules, communication styles, workload, and collaboration habits. We wanted to build a "CAT" that could sit with a group of students and help them work better together, while still keeping instructors in control.

What it does

CAT helps both instructors and students throughout the full team-collaboration lifecycle.

For instructors, CAT provides:

A dashboard with roster setup, KPI snapshots, and freshness alerts for derived team data
Lightweight roster creation before students complete deeper onboarding
AI-generated student collaboration profiles based on questionnaire responses
Deterministic team formation that balances skill coverage, schedule overlap, communication diversity, leadership distribution, and growth fit
Transparent team rationale, score breakdowns, and risk flags so placements are explainable
Instructor override tools to move students, analyze swaps, reroute members, and repair out-of-bounds teams

For students and teams, CAT provides:

A detailed onboarding questionnaire covering availability, strengths, growth areas, workload, communication habits, leadership preference, prior experience, and goals
A team workspace with generated charter, suggested role rotation, and kickoff checklist
AI tools to rewrite messages, summarize meeting notes into action items, and generate team charters
A Team Copilot that can preview and execute bounded actions like message rewriting, meeting planning, meeting follow-up task creation, and weekly team status summaries
Participation signals and disengagement alerts so instructors can spot teams that may need support
Google sign-in for identity and optional Google Calendar OAuth for scheduling invites and Meet links
Good-standing team badge support with Solana devnet-ready credential references

How we built it

We built CAT with Next.js, React, TypeScript, Tailwind CSS, reusable Radix-based UI components, and Zod-validated API routes. Gemini powers profile generation and AI text workflows like charters, summaries, rewrites, rationale generation, and coaching suggestions. For team formation, we combined AI assistance with deterministic heuristic scoring so the system stays transparent and auditable instead of becoming a black box. We also added a mock-first repository layer and fallback responses so the entire demo remains reliable even when cloud credentials are unavailable.

Challenges we ran into

One of the biggest challenges was designing an AI-assisted system that instructors could trust. We did not want a black-box sorter, so we built transparent scoring, visible risk flags, and override workflows. Another challenge was reliability: generative APIs can fail or rate-limit, so we added schema validation and deterministic mock fallbacks to keep the demo stable. We also had to separate Google identity from optional Calendar consent so students remain in control of what they connect.

Accomplishments that we're proud of

We're proud that CAT goes beyond simple team assignment. It supports onboarding, profiling, transparent matching, instructor intervention, meeting planning, action tracking, communication support, and team health monitoring in one workflow. We're also proud that the system already bridges multiple surfaces: instructor tooling, student tooling, Gemini-powered AI workflows, Google integrations, and badge-ready credentials.

What we learned

We learned that the best education AI should augment, not automate. Instructors need visibility and override power, and students need lightweight support that actually reduces team friction. We also learned that combining deterministic systems with bounded AI actions creates a much more trustworthy product than relying on generation alone.

What's next

Next, we want to connect CAT to real classroom and LMS signals, improve live participation analytics, expand instructor intervention workflows, deepen the badge system, and turn Team Copilot into a persistent collaboration companion for student teams over an entire semester.