Open-Visa-Pathway

• 

  An example pathway showcasing when a Nigerian student accounts for their full stay in the target country .

• 

  An example pathway showcasing when an Indian student financial budget doesn't account for their full stay in the target country.

• 

  AI chatbot + RAG kb to verify links in real time and unofficial sources.

• 

  Demo page outlining problem statement, goal and innovative functionalities for this hackathon.

• 

  AI chatbot response supported with real time links.

Inspiration

International students face confusing, fragmented visa + university requirements. I wanted a tool that turns “visa chaos” into an action plan.

What it does

Open-Visa-Pathway helps users:

• Wizard → Roadmap: generates a corridor timeline (phases + steps) with a downloadable CSV.
• Document explainer: paste a scary email/requirements → get a plain-English summary + checklist + “UNKNOWNs”.
• Collaborator helper: drafts outreach messages + suggests time slots for cross-time-zone coordination.

How I built it

• Frontend: Streamlit (tabs + sidebar navigation)
• AI: OpenAI chat completions for structured outputs
• RAG: knowledge-base markdown chunks → embeddings → cosine similarity retrieval → “sources used” shown in UI
• State: Streamlit session_state to keep chat history and wizard progress

Challenges I ran into

• Making the “Document explainer” return strict JSON reliably
• Keeping retrieval transparent (show sources, avoid hallucinating missing details)
• UX: routing between sidebar nav and tabs without confusion

What I learned

• How to design an LLM feature with guardrails (UNKNOWNs, sources, structured schema)
• How to implement lightweight RAG for explainability
• Streamlit layout patterns that improve demo clarity quickly

What’s next for Open-Visa-Pathway

• Add country-specific official link packs per corridor
• Add a “save/export” flow for checklists and drafted outreach
• Improve validation + testing around JSON outputs

Built With

• openai
• python
• streamlit