Industry
EdTech
Case Study
AI Learning Platform
Teachers often don't know a student's in trouble until a test proves it. At brigit.dev I worked on the backend and AI suggestions. Who needs help today, and what can a teacher assign before the next class?
Project Overview
Teachers often don't know a student is struggling until a test proves it. By then it's late. At brigit.dev I worked on a product that flags at-risk learners early and suggests activities teachers can assign without learning a whole new system.
Teachers, school admins, and training managers all see the same student data, just from different angles. I built across all of it: the teacher views, the admin views, the backend, and the AI suggestion layer.
School product that shows which students are falling behind, and suggests what teachers can do about it today.
Services
- UX/UI Design
- Web Development
- Product Strategy
Key Deliverables
- Responsive Design
- Custom Development
- User Experience Improvements
Platform Type
Web Platform
Tech Stack
- TypeScript
- Vite
- React
- NestJS
- i18n
- PostgreSQL
- WebSocket
The Challenge
Class averages hide individual problems. One student quietly falling behind looks fine on a gradebook summary. Teachers are already overloaded, they won't open another dashboard they'll ignore by week two.
The product had to answer a simple question: who needs help today, and what can I do about it in the next ten minutes?
Approach
I started from what a teacher does between classes, not from a list of AI features. Open the app. See which students dropped off. Assign something or send a reminder. Done.
At-risk alerts are the main signal, not buried in a chart somewhere. AI suggestions are things a teacher can actually assign: an activity, a quiz, a nudge. If a suggestion isn't useful, they skip it. No black box.
Live classroom data comes through WebSockets. Multiple languages supported because the schools using it weren't all English-only.
The Solution
Teachers see each student's progress, which modules they've finished, where they're stuck, whether they're at risk. Not a class average. A person.
The AI part suggests interventions: "this student hasn't touched module 3, here's a short activity." Teachers assign it in one click or ignore it. No new process to learn.
Admins and training managers see enrollment, class-level progress, and which groups need attention, same data, different questions.
Key Challenges
-
AI suggestions teachers would actually assign, not impressive demos they click once and forget
-
showing the same student data clearly to three different roles without it getting contradictory or duplicated
-
shipping the whole thing, live classes, admin tools, multilingual UI, AI layer, not just one polished screen
The Outcome
Teachers, admins, and training managers work from the same student data. At-risk kids get flagged before they fully fall off. Teachers get suggestions they can act on in minutes, not reports they read once a semester.
The whole product shipped. It's not a prototype with one good demo flow.
Context
Built at brigit.dev. Three roles, one data source. That consistency mattered as much as any single feature, a teacher and an admin looking at the same student should see the same story.
The AI only works if teachers trust it. That meant specific suggestions, easy to dismiss, never pretending to know better than the person in the classroom.