The Problem
Academic publishing workflows are usually shaped by unclear requirements, many roles, and long feedback loops. I wanted to build a system that could handle those realities while also being comfortable to develop with AI agents.
The Approach
I chose a stack that fits both the product and the way I work:
FastAPIfor the backend because the project needed text processing, API orchestration, and Python-friendly AI workflows.Next.jsfor the frontend because it fits well with TypeScript, React, and AI-assisted UI iteration.shadcnfor UI consistency and fast component work.Supabase,Vercel, andHugging Face Spacesto keep the deployment story practical across the stack.
What It Demonstrates
ScholarFlow is the project where I moved from “AI helps me write code” to “I design workflows that help AI and humans make better decisions.”
It covers:
- product and workflow thinking
- full-stack implementation
- requirement clarification through agent conversations
- git-history-driven iteration
- my shift from GUI-first AI tools to CLI agents and skill-based workflows
This is the project I would point to when explaining how I approach complex product work end to end.