UniMate

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

Inspiration

University does not usually fail because students have no to-do list. The harder problem is deciding what deserves attention when assignments, classes, errands, money, meals, work, friends, and rest all compete at once.

That problem becomes worse for students living independently for the first time. They have less structure, limited budgets, unfamiliar schedules, and dozens of small decisions to make every day. We built UniMate to turn that mental overload into one clear next move.

What it does

UniMate is a voice-first prioritization companion represented by a small penguin.

A student can speak or type a messy brain dump such as:

“I have a lab soon, groceries to buy, an assignment due tomorrow, and only 30 minutes free.”

UniMate extracts the actual tasks and constraints, checks the student’s available time, timetable, deadlines, effort, money, food, and rest, then returns:

• One task to do now
• What comes next
• What can wait
• Why the recommendation changed
• What delaying each task could cost

Students can change their context with follow-ups such as “I only have 25 minutes.” UniMate immediately reranks the plan instead of simply displaying the new information.

UniMate works as both a SwiftUI iPhone app and a Chrome side-panel extension. The Chrome extension can also scan a course page for deadlines when the student explicitly requests it.

How we built it

The iOS app is built with SwiftUI and includes voice capture, typed input, task details, contextual reranking, schedules, decision history, reminders, a focus timer, and a Live Activity.

A Node.js and Express API receives text or audio and sends live requests to Snowflake. Audio is stored in a Snowflake stage and processed with AI_TRANSCRIBE. Snowflake stored procedures use AI_COMPLETE to extract structured tasks and create the plan.

The ranking engine combines deadlines, available time, effort, importance, money at risk, basic needs, and schedule conflicts. Every plan, task, capture, action, and ranking decision is stored in Snowflake so the result remains explainable and auditable.

ElevenLabs gives UniMate a natural spoken voice. We also built a Manifest V3 Chrome extension using JavaScript, HTML, and CSS that connects to the same API and student data.

Challenges

The hardest challenge was turning unstructured speech into reliable decisions. A transcript may contain tasks, deadlines, emotions, money constraints, and schedule information in a single sentence.

We also had to ensure context changed the actual ranking rather than becoming another filter. A task that makes sense with 70 minutes available may be impossible with 25, so UniMate must explain what moved and why.

Voice introduced additional failure cases involving permissions, audio formats, transcription latency, and network access. We built on-device transcription and deterministic ranking fallbacks so the student’s original input is preserved when a service fails.

Finally, each hackathon pivot changed the product itself. The university-student pivot introduced budgets, classes, food, rest, and independent living. The context pivot made available time, deadlines, effort, and importance central to every recommendation.

Accomplishments

• Built working SwiftUI and Chrome experiences around one backend
• Connected live Snowflake storage and Cortex processing
• Created a contextual ranking engine instead of another task list
• Added voice input and spoken recommendations
• Made every recommendation explainable through evidence and history
• Preserved real student input through network and transcription failures
• Added focus timers, reminders, schedules, Live Activities, and course-page scanning

What we learned

We learned that prioritization is not the same as sorting by deadline. A useful recommendation must account for what fits now, what becomes more expensive to delay, what conflicts with the schedule, and what the student realistically has the energy or money to complete.

We also learned that AI is more useful when it produces a concrete decision that can be inspected. Snowflake is not simply storing UniMate’s output. It holds the tasks, context, history, and procedures that create and explain the ranking.

What's next

Next, we want to add opt-in calendar synchronization, collaborative study-session planning, stronger course-platform imports, and delivery-aware meal planning. The long-term goal is for every student to have a practical companion that understands their changing day and helps them take the next step.

What we learned

What's next for UniMate

Built With

• github
• snowflake