Skip to content

odominguez7/meetMIT

BranchesTags

Repository files navigation

meetMIT

real connections, agentic precision.

AI-powered campus connection platform that turns free moments into meaningful in-person meets. Big Five personality matching. Agent-assisted scheduling. Feedback loops that learn. Built for MIT and Harvard.

License: MIT Next.js PostgreSQL TypeScript PRs Welcome

Open source. Contributions welcome.

The problem

Campus life is full of people you should know but never meet.

You walk past 500 people a day at MIT. Some share your research interests. Some are building the startup you'd join. Some just want a study partner who actually shows up. But the default social discovery tools are either too passive (hoping you bump into someone) or too performative (LinkedIn, mixers, forced networking events).

The result: Most meaningful campus connections happen by accident, not by design.

The solution

meetMIT uses personality science and AI agents to create high-quality, low-friction micro-connections.

You:       Complete a 2-minute "Vibe Check" (10 questions)
meetMIT:   Maps your answers to Big Five (OCEAN) personality traits
meetMIT:   Finds people with 70%+ affinity match on campus
meetMIT:   Creates "Sparks" — 15-60 minute micro-meets (coffee, study, walk, brainstorm)
You:       Show up. Connect. Rate the experience.
meetMIT:   Learns from feedback. Next match is better.

No swiping. No profiles to curate. No pressure. Just show up.

How it works

Affinity matching

Users complete a 10-question "Vibe Check" mapped to the Big Five (OCEAN) personality model:

Trait Questions What it measures
Openness 1-2 Curiosity, creativity, openness to new experiences
Conscientiousness 5-6 Organization, reliability, follow-through
Extraversion 3-4 Energy from social interaction, spontaneity
Agreeableness 7-8 Empathy, cooperation, trust
Neuroticism 9-10 (reversed) Emotional stability, stress response

Answers (1-5 scale) are normalized to 0-1 vectors. Matching uses cosine similarity between OCEAN vectors:

  • 70%+ similarity = core match (shared wavelength)
  • Strategic complements = people who balance your weak spots (surfaced separately)

Spark types

Seven types of micro-meets, each designed for a different energy level:

Spark Duration Vibe
Study Sync 30-60 min Parallel focus work
Coffee Clash 15-30 min Quick 1:1 conversation
Lunch Link 45-60 min Meal + deeper conversation
Walk Wave 15-30 min Walking meeting
Homework Hack 60 min Collaborative problem-solving
Recess Rush 15 min Quick energy break
Blitz Brainstorm 30 min Rapid idea exchange

Feedback loops

After every meet, both people rate the experience (1-5) and indicate repeat interest. This data feeds back into the matching algorithm. The system gets smarter with every connection.

Architecture

┌─────────────────────────────────────────────────────────────┐
│                     Browser (React 18)                       │
│                                                             │
│  Landing · Onboarding · Sparks Feed · Offers Board          │
│  Insights Dashboard · Trust & Safety · Settings             │
│  AI Studio (Agent Simulation Viewer)                        │
└───────────────────────────┬─────────────────────────────────┘
                            │ REST API (24 endpoints)
┌───────────────────────────▼─────────────────────────────────┐
│                    Next.js 14 (App Router)                    │
│                                                             │
│  ┌─────────────┐  ┌──────────────┐  ┌───────────────────┐  │
│  │  Zod         │  │  Affinity    │  │  Agent Runtime    │  │
│  │  Validation  │  │  Engine      │  │  (7 personas)     │  │
│  │              │  │              │  │                   │  │
│  │  All inputs  │  │  OCEAN       │  │  Registration     │  │
│  │  validated   │  │  cosine      │  │  Interaction      │  │
│  │  at boundary │  │  similarity  │  │  Observability    │  │
│  └──────────────┘  └──────────────┘  └───────────────────┘  │
│                                                             │
│  ┌──────────────────────────────────────────────────────┐   │
│  │           Prisma ORM + PostgreSQL 16                  │   │
│  │  users · sparks · commitments · feedback · offers     │   │
│  │  affinity_profiles · trust_reports · blocks · metrics │   │
│  └──────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────┘

Tech stack

Layer Technology
Framework Next.js 14 (App Router, standalone output)
Language TypeScript 5.7 (strict mode)
Database PostgreSQL 16 + Prisma ORM
UI React 18, Tailwind CSS, Framer Motion
Validation Zod (all API inputs validated at boundary)
Auth .edu email verification (MIT/Harvard only)
Deployment Docker (multi-stage, non-root) + Google Cloud Run

Pages and features

Page What it does
/ Landing page with hero, how it works, spark types
/onboarding 4-step wizard: email verification, vibe check, availability, confirmation
/sparks Browse and filter spark opportunities with affinity scores
/sparks/[id] Spark detail, commit to attend, post-meet feedback
/offers Value exchange board ("I want X" / "I offer Y")
/insights Campus-level aggregate analytics (no PII exposed)
/trust Report safety incidents, block users
/settings Automation level, notifications, visibility, account actions
/ai-studio-room Agent simulation viewer (7 AI personas interacting)

API reference

Users

Method Endpoint Description
POST /api/users/verify-edu Verify .edu email, create/return user
POST /api/users/onboard Save preferences and availability modes

Affinity

Method Endpoint Description
POST /api/affinity/profile Upload Big Five OCEAN personality scores

Sparks

Method Endpoint Description
GET /api/sparks/recommendations Personalized spark feed with affinity matching
POST /api/sparks/create Create a new spark opportunity
POST /api/sparks/[id]/commit Commit to attending a spark
POST /api/sparks/[id]/feedback Post-meet rating + repeat interest

Offers

Method Endpoint Description
GET /api/offers/create List unfulfilled offers
POST /api/offers/create Post a value exchange offer

Trust & Safety

Method Endpoint Description
POST /api/trust/report Report a safety incident
POST /api/trust/block Block a user from future matches

Insights

Method Endpoint Description
GET /api/insights/campus Aggregate connection metrics (privacy-safe)

Agents

Method Endpoint Description
GET /api/agents List all registered agents
POST /api/agents/register Register a new agent
GET /api/agents/[id] Agent details + reports
PATCH /api/agents/[id]/status Pause/resume agent
POST /api/agents/[id]/interact Send task to agent
GET /api/agents/[id]/activity Agent activity log (paginated)
POST /api/agents/[id]/report Report agent for abuse

Admin

Method Endpoint Description
GET /api/admin/observability Agent observability summary
GET /api/admin/activity-log System-wide activity log
GET /api/health Server health check
GET /api/health/db Database connectivity check

Quick start

Local development

git clone https://github.com/odominguez7/meetMIT.git
cd meetMIT
npm install

# Start Postgres
docker compose up db -d

# Configure environment
cp .env.example .env

# Push schema and seed demo data
npx prisma db push
npm run db:seed

# Run dev server
npm run dev

Open localhost:3000. The seed creates 7 demo personas, 8 sparks, and 5 offers.

Deploy to Cloud Run

# Build and push container
gcloud builds submit --tag gcr.io/PROJECT_ID/meetmit

# Deploy
gcloud run deploy meetmit \
  --image gcr.io/PROJECT_ID/meetmit \
  --platform managed \
  --region us-east1 \
  --set-env-vars DATABASE_URL="YOUR_CLOUD_SQL_URL" \
  --set-env-vars API_SECRET="$(openssl rand -hex 32)"

Environment variables

Variable Required Description
DATABASE_URL Yes PostgreSQL connection string
API_SECRET Prod only Strong random string for token signing
NEXT_PUBLIC_APP_URL No Frontend URL (default: http://localhost:3000)
DB_USER / DB_PASSWORD / DB_NAME / DB_SOCKET No Cloud SQL Unix socket alternative

Database schema

10 models across identity, connections, trust, and analytics:

users ──────────── affinity_profiles (1:1, OCEAN scores)
  │
  ├── sparks ──────── spark_commitments (who's attending)
  │                    spark_feedbacks (post-meet ratings)
  │
  ├── offers (value exchange: "I want X / I offer Y")
  │
  ├── trust_reports (safety incidents)
  │
  ├── blocks (user-to-user blocking)
  │
  └── user_avail_modes (casual / study_only / founder_mode / zen_only)

connection_metrics (daily aggregates, no PII)
activities (system-wide event log)

Project structure

app/
  page.tsx                 Landing page (hero + how it works)
  layout.tsx               Root layout + navigation
  onboarding/page.tsx      4-step onboarding wizard
  sparks/page.tsx          Browse spark opportunities
  sparks/[id]/page.tsx     Spark detail + commit + feedback
  offers/page.tsx          Value exchange board
  insights/page.tsx        Campus analytics dashboard
  trust/page.tsx           Safety controls + reporting
  settings/page.tsx        User preferences
  ai-studio-room/page.tsx  Agent simulation viewer
  api/                     24 REST endpoints (see API reference)

components/                Reusable React components
lib/
  db.ts                    Prisma client singleton (Cloud SQL support)
  schemas.ts               Zod validation schemas
  affinity.ts              OCEAN matching + vibe check scoring
  api.ts                   Response helpers
  agents-store.ts          Agent registry + activity log
  ai-simulation.ts         Agent simulation engine

prisma/
  schema.prisma            Full data model (10 models)
  seed.ts                  7 demo personas + 8 sparks + 5 offers

Contributing

meetMIT is open source and built for collaboration. Whether you're at MIT, Harvard, or anywhere else -- if you care about making real human connections easier, you're welcome here.

Good first issues

  • Add authentication -- integrate NextAuth.js with .edu email verification
  • Build the spark creation form -- replace the modal placeholder with a real form
  • Add tests -- unit tests for affinity matching, integration tests for API routes
  • Expand to other campuses -- add campus enum values and location data
  • Improve matching -- experiment with different similarity algorithms beyond cosine
  • Mobile polish -- responsive improvements for the sparks feed and onboarding flow
  • Add notifications -- email or push notifications for spark invites and reminders

How to contribute

# 1. Fork the repo
# 2. Clone your fork
git clone https://github.com/YOUR_USERNAME/meetMIT.git
cd meetMIT

# 3. Set up locally (see Quick Start above)
npm install
docker compose up db -d
cp .env.example .env
npx prisma db push
npm run db:seed
npm run dev

# 4. Create a branch
git checkout -b my-feature

# 5. Make changes, commit, push, open a PR

Code style

  • TypeScript strict mode -- no any unless absolutely necessary
  • Zod validation on every API input
  • React components in PascalCase, files in kebab-case
  • Keep it simple. If it doesn't need abstraction, don't abstract it

The bigger picture

Campus connections shouldn't depend on luck.

The people who change your life at MIT aren't the ones in your section or your dorm. They're the ones you almost never meet because you're in different buildings, different schedules, different orbits. meetMIT closes that gap with personality science, not algorithms optimizing for engagement.

Every connection is real. Every meet is in person. Every match gets better because the system learns from what actually works. Not likes, not swipes, not followers. Just two people who showed up and had a conversation worth having.

Real connections. Agentic precision.

Built by Omar -- MIT Sloan '26

MIT License. Open source. Contributions welcome.

About

AI-powered campus connections for MIT and Harvard. Big Five personality matching, micro-meets, feedback loops. Next.js 14 + PostgreSQL + Prisma.

Topics

react docker open-source typescript ai mit matching nextjs postgresql campus personality prisma

Resources

Readme

License

MIT license
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages