• 

  Sign up Page

• 

  Job Description Page

• 

  AI chat bot view

• 

  Call View

About JobJam

Inspiration:
I noticed a gap in most career prep tools: while AI can help users improve, they often lack human connection. Practicing with a real person — someone with similar goals — makes the experience more engaging, realistic, and effective. I wanted to create something like an "Omegle × LinkedIn Hybrid," helping people improve public speaking, interview skills, and networking, all without needing existing connections or social capital.

What I Learned:

• How to integrate real-time video calls in a mobile app using Agora.
  
• How to use vector embeddings for semantic matching with Hugging Face models.
  
• How to manage anonymous authentication and real-time data with Firebase and Firestore.
  
• How to design a system for privacy-first matchmaking while maintaining responsiveness.
  

How I Built It:

1. Users sign in anonymously with Firebase Authentication — no account required.
   
2. They enter a job description, which is converted into a vector embedding using the Hugging Face intfloat/e5-large-v2 model.
   
3. The embedding is stored in Firestore, and users are placed in a matchmaking queue.
   
4. I calculate cosine similarity between embeddings to find compatible matches. If the similarity passes a threshold, users are paired in a real-time video call via Agora, with a token server deployed on Render.
   
5. During calls, users can mute audio, disable video, or leave. They also have access to JobBot, an AI assistant powered by Gemini, which can provide career guidance, simulate interview prompts, and answer questions in real time.
   

Challenges Faced:

• Ensuring fast and reliable matchmaking without revealing any personal information.
  
• Integrating real-time video with matchmaking logic in a seamless experience.
  
• Managing asynchronous calls to Hugging Face and storing embeddings efficiently.
  
• Designing the system so users can interact anonymously yet meaningfully.
  

Next Steps:

• Adding local audio recording, transcription, and evaluation with AssemblyAI.
  
• Improving JobBot’s personalization to provide more tailored career advice.
  
• Making the app scalable to handle larger matchmaking queues while keeping low latency.

Built With

• agora
• firebase
• gemeni
• render
• swift
• swiftuirtc
• xcode