About the Project
VisaStrategy AI was inspired by how confusing U.S. immigration planning can be for talented workers, founders, and small teams. Many people do not know whether O-1 or H-1B is the better path, what evidence matters, or where their profile is weak.
We built a Streamlit app powered by LangGraph and a local RAG corpus. The system classifies the likely visa path, checks eligibility signals, recommends evidence, identifies risks, and cites sources from legal, evidence, and precedent collections.
We learned how to combine workflow graphs, retrieval, structured outputs, and offline embeddings into a working legal-tech prototype. The biggest challenge was dealing with blocked government pages, missing PDFs, and making the app run without an API key.
Our confidence score is a simple demo heuristic:
[ score = \frac{\text{satisfied criteria}}{\text{required criteria}} ]
VisaStrategy AI is not legal advice, but it gives users a faster, clearer first-pass strategy before talking to an immigration lawyer.
Built With
- beautiful-soup
- chromadb
- github
- langchain
- langgraph
- prompts
- pypdf
- python
- streamlit
Log in or sign up for Devpost to join the conversation.