FairLens

Problem

Grant and scholarship review is slow, inconsistent, and prone to bias. Human reviewers apply different standards across applications, miss relevant criteria, and can’t process volume at scale. Applicants receive little actionable feedback.

Approach

Built a Turborepo monorepo with a Next.js frontend, FastAPI backend, and a Google ADK multi-agent system with 7 specialized agents. Each agent handles a distinct phase of the review pipeline — from intake and eligibility screening through scoring, comparison, and recommendation generation.

Features

• 7 specialized agents — intake, eligibility, scoring, comparison, feedback, escalation, and orchestrator agents built on Google ADK
• Multi-provider LLM routing — LiteLLM abstraction layer switching between providers based on cost and capability requirements
• Full SaaS stack — Clerk authentication, Stripe billing, Prisma ORM with PostgreSQL
• End-to-end testing — Playwright test suite covering critical user flows
• Railway deployment — containerized deployment with environment-based configuration

Results

13 planning documents produced during architecture phase. Complete multi-tenant SaaS architecture with role-based access, audit logging, and multi-currency billing.