Built by: Emily Yan, Edison Cai, Vincent Bei, and Lily You

Networking and the ability to build connections is one of the most decisive factors in the job search. However, it can be hard to manage thorough preparation for networking events atop a heavy courseload, personal projects, friends, family, and more. MeetBetter reduces this burden, making events more approachable and accessible for everyone.

MeetBetter uses facial and voice recognition to find and scan a stranger's LinkedIn autonomously, and delivers the perfect amount of context kick-start conversations right away. Instead of forcing users to pull out their phone, search LinkedIn manually, and read long profiles mid-conversation, streamlines your networking experience in real-time.

Using a laptop’s built-in webcam (as a stand-in for wearable assistants like Meta Glasses), MeetBetter detects a face and displays a small, anchored prompt card beside the person, providing a personal overview, quick talking points and conversation starters based on their portfolio. This keeps networking simple, fast, and accessible.

It’s obvious the job market is “cooked.” Networking has become the hidden filter for jobs, creating a barrier that impacts stability, independence, and opportunities. Making networking more accessible matters. MeetBetter focuses on low-effort accessibility by:

• reducing cognitive load (short prompts instead of long profiles)
• keeping attention on the person (no phone-scrolling)
• using a fast, readable UI (minimal text, structured layout)

1. Open MeetBetter and your laptop camera turns on.
2. When you look at someone, MeetBetter detects their face and places a small prompt card beside them on-screen.
3. The card shows minimal, useful context (name + a few quick talking points) so you can start a conversation without pulling out your phone.
4. If you say a name out loud, MeetBetter can lock onto that person briefly to keep the card stable and reduce “jumping” between people.
5. After the interaction, you can use the same quick info to remember what to follow up on (prototype/demo content may be a placeholder or locally stored).

Note: This is a prototype UI/UX demo. Any “profile” content shown is a placeholder or stored locally for the hackathon.

Backend (Python)

• FastAPI — lightweight API layer to connect modules (vision ↔ voice ↔ UI signals)
• MediaPipe — real-time face detection + anchor point tracking
• OpenCV — webcam capture + drawing the on-screen anchored card UI
• Vosk (offline speech recognition) — converts live microphone audio into text commands/names
• NumPy — embeddings + similarity scoring for local face matching (prototype DB)
• Gemini (Google LLM API) — contextual reasoning and dynamic conversation prompts generated from profile data
• InsightFace — high-accuracy facial recognition and embedding extraction for identity matching