ArrivalMate

A Gemini + MongoDB agent that helps newcomers navigate their first month in Canada.

Comment

Inspiration

Moving to a new country can feel overwhelming, especially during the first few weeks. Newcomers and international students often struggle with scattered information about banking, transit, healthcare, housing safety, government documents, school setup, and scam prevention.

As an international student in British Columbia, I saw how stressful and confusing the first month in Canada can be. Small but important tasks quickly become overwhelming when everything happens at once.

I wanted to build something that goes beyond a simple chatbot. Instead of only answering questions, I wanted an AI agent that could organize priorities, track progress, adapt over time, and create a structured onboarding experience for newcomers.

That idea became ArrivalMate AI.

What it does

ArrivalMate AI is an AI-powered newcomer support agent that helps users navigate their first 30 days in Canada.

Users complete a personalized onboarding flow with details such as:

city arrival status onboarding needs school or work setup priorities

The system then generates:

prioritized onboarding tasks grouped timelines (Today, This Week, This Month) scam prevention guidance local support resources adaptive planning recommendations progress tracking persistent onboarding memory

Users can also:

mark tasks complete refine plans dynamically revisit saved onboarding snapshots track onboarding history over time

ArrivalMate AI focuses on reducing overwhelm and helping newcomers make safer and more informed decisions during a critical transition period.

How we built it

ArrivalMate AI was built using:

Next.js TypeScript Tailwind CSS MongoDB Atlas Gemini-style planning workflows Vercel deployment

MongoDB stores:

user profiles onboarding plans progress tracking saved plan snapshots refinement history agent activity logs

The AI planning layer generates structured onboarding workflows instead of conversational-only outputs. Plans are grouped by urgency and continuously refined as the user’s situation changes.

The system also includes resilient fallback planning support, allowing the app to remain functional even if live AI generation is temporarily unavailable.

Challenges we ran into

One of the biggest challenges was balancing realism, reliability, and usability.

AI-generated outputs can sometimes become inconsistent or overly generic, so I had to build validation and normalization logic to keep plans structured and context-aware.

Another challenge was designing onboarding recommendations that respected different user statuses. For example:

visitors should not automatically receive work-related onboarding tasks students may require school support and work-eligibility guidance workers need payroll, banking, and transit setup priorities

I also focused heavily on graceful fallback behavior. During testing, Gemini quota limitations occasionally occurred, so I built resilient demo-mode support to ensure the application remained stable and demo-ready while preserving the onboarding workflow.

Accomplishments that we're proud of

Built a functional AI planning agent instead of a simple chatbot Created adaptive onboarding workflows with persistent memory Implemented MongoDB-backed progress tracking and saved plan snapshots Added scam prevention guidance tailored to newcomer risks Built resilient fallback support so the app remains usable even if live AI generation is unavailable Successfully deployed the full application publicly on Vercel

What we learned

This project reinforced how important workflow design and clarity are when building AI systems.

I learned that the most useful AI experiences are not always the most technically complicated. The strongest experiences come from:

solving a real problem structuring information clearly maintaining persistent memory adapting over time reducing user stress

I also gained deeper experience with:

MongoDB data modeling structured AI planning flows frontend state management deployment workflows resilient fallback architecture hackathon-focused product development

What's next for ArrivalMate

Future improvements could include:

multilingual onboarding support calendar and reminder integration verified newcomer datasets real-time transit integration budgeting tools school and workplace integrations AI-assisted document readiness workflows

The long-term vision is to create a practical AI onboarding companion that helps newcomers feel more confident, informed, and supported during one of the biggest transitions of their lives.

Built With

Share this project:

Updates