FULL-STACK RAG PRODUCT

Borderless AI Copilot

PROBLEM

Immigration rules are overly complex, constantly changing, and vary drastically across 70+ countries. Finding out which visas you qualify for requires hiring expensive lawyers, and most free online assessments are lead-generation scams.

SOLUTION

Built an evidence-grounded immigration copilot combining deterministic eligibility scoring with RAG (Retrieval-Augmented Generation).

Deterministic Rules Engine: Calculates exact points for systems like Canada's Express Entry or Germany's Chancenkarte based on hardcoded mathematical rules, not AI guesswork.
RAG Chatbot: Vectorizes official government documents using Neon pgvector. The AI answers follow-up questions purely based on retrieved, official .gov and .gc.ca sources.
What-If Simulator: Allows users to tweak their profile (e.g., increasing English score) and instantly recalculates global visa eligibility.

ARCHITECTURE: Hybrid Scoring & RAG

→ User Profile Input

→ Deterministic Engine (Exact points calculated)

→ Neon pgvector retrieves official Gov docs

→ LLM injects context & explains next steps

✓ Grounded Answer + Simulator Generation

RESULT

Successfully modeled 1,454 distinct visa pathways across 70+ countries into a unified mathematical scoring engine.

Launched a complete B2C SaaS product featuring Stripe monetization, Clerk authentication, and a scalable Next.js 14 architecture.

KEY METRICS

1,454 Visa Pathways Modeled

70+ Countries

Deterministic + RAG Hybrid

Full SaaS Architecture

TECH STACK

Next.js 14Neon (pgvector)ClerkStripeOpenAI