AI Immigration Lawyer Agent

Inspiration

We need an AI chatbot that can give personalized immigration recommendations based on user data and frequent immigration changes.

What it does

• AI chatbot that answers user immigration questions. (E.g, personalized pathway recommendations.)
• Dynamic CRS scores calculation based on recent updates.

How we built it

• AI multi-agent: Crew.ai agents ( CRS rules analyst agent to check recent changes, CRS score calculation agent, and a flow for dynamic switching)
• Context engineering: An openrouter Gemini-2.0 flash model with user profile, uploaded docs, CRS score, conversational history, and Talivy webtool to search.
• OCR pipeline: Landing.ai to detect document layout, document type, and text extraction.
• MongoDB: For unstructured document storing.
• FastAPI: Backend.
• React(vite) & Typescript: Frontend.

Challenges we ran into

• Proper context engineering to reduce LLM hallucination.
• Parameter tuning for deterministic results.
• Security to store sensitive immigration data.

Accomplishments that we're proud of

• Learned new technologies.
• Team effort from ideation to MVP within 24hours.

What we learned

• The complexity of natural language document extraction.
• Utilizing AI agent flow and tool capability for dynamic and robust decision making.

What's next for AI Immigration Lawyer Agent

• Support different languages document extraction. ( eg, Persian, Chinese, Bangla )
• Enhanced security to save sensitive data on the user's end.
• Notification for deadline/document expiration.
• Chat history for improved LLM memory.
• Fine-tuned model for open-ended chat.
• Voice model for people with disabilities.

Built With

• crew.ai
• fast.api
• gemini
• landing
• mongodb
• react.js
• typescript