Senac Essay Contest
[Challenge]How many essays come in when an institution like Senac opens a contest? Five hundred? Five thousand? Fifty thousand? Nobody knows until the deadline closes. And without that number, there's no way to size the problem. How many graders to hire? How long will each one need? What's the total cost? When can the results go out? That's why Senac's first essay contest had to be made possible by AI.
[Solution]We built a system that solves all that unpredictability with AI. The AI here doesn't give each essay its final grade. We wouldn't trust it with that yet. It filters. Whatever the volume of entries, the eighty best always reach the human jury. Eighty essays we know how to budget and schedule.
Each essay is scored on the 5 criteria Senac selected, by two different LLMs so one keeps the other honest. When Claude and ChatGPT disagree too much, the system re-evaluates and lands on the right score. All with filters to prevent prompt injection, calibrated to verify itself.
The human jury keeps the part that was always theirs: choosing the winner among the best.
AI gives you the predictability to act without logistics turning into an obstacle.
The flow from essay submitted to winner decided. AI filters; a human chooses.
Well-calibrated systems don't take the human out of judging. They take the human out of screening.
Thousands of essays stop being loose text and become a map: what this generation in Rio believes, hopes for, and ignores.
- Client
- Senac RJ
- DATE
- December 2025
- ACCOUNT MANAGEMENT
- Tainá Scandiuzzi
- DEVELOPMENT
- Fernando Feitosa
- AI STRATEGY
- Felipe Pacheco
- CALIBRATION
- Felipe Pacheco