Inspiration
Every year, hundreds of thousands of Moroccan students face a critical fork in the road the days after receiving their baccalauréat results. They have hours to choose a path that will shape the rest of their lives, often with no counselor available, no personalized guidance, and overwhelming pressure from family and society. Traditional orientation tools ask cold, bureaucratic questions: "What's your BAC branch? What's your average?" as if a grade is who you are.
We were inspired by the idea that a good counselor doesn't ask for your grades. They ask who you are. Tawjihi was born from this insight: an AI companion that speaks to Moroccan students in their own language-Darija-like a smart older friend, not a government form.
What it does
Mowajihi AI (Tawjihi) is an AI-powered orientation companion for Moroccan students before and after the Baccalauréat. Instead of rigid quizzes or forms, it:
- Starts a natural conversation in Darija: opening with "Khberni 3la rasek" to build trust before any academic questions.
- Silently builds a personality profile across a 6–8 message conversation, inferring work style, motivation type, risk tolerance, social preference, and practical constraints.
- Reveals a personalized career roadmap: around 3 ranked career paths, each explained in the student's own words and language, with a clear "why this fits you."
- Recommends real Moroccan schools: filtered by city and public/private preference, with actual admission requirements and filières (ENSA, ENCG, Faculté de Médecine, etc.).
How we built it
- Frontend: Angular 17+ with standalone components and PrimeNG UI library.
- AI: Gemini 2.0 Flash via direct REST API calls.
- Backend: use of Firebase for authentication and data storage.
Challenges we ran into
- Prompt engineering for language switching: Getting Gemini to reliably stay in Darija without slipping into Modern Standard Arabic.
- Stateless AI, stateful conversation: The Gemini API has no built-in session memory, so we had to pass the entire growing conversation history on every call and persist it to Firestore to survive page refreshes.
Accomplishments that we're proud of
- A chat experience that feels genuinely warm and human.
- A hidden personality profiling engine that builds a structured
StudentProfileobject (traits, motivation, risk tolerance, and work style) purely from casual conversation, with no forms. - A real, contextual database of Moroccan schools with actual admission criteria, not generic placeholder data.
What we learned
- How we divide work between us and keep the context of the project between all our IDEs
- The most important file in an AI product is the system prompt. More time spent on the prompt yields better results than any model switch.
- Moroccan students are deeply underserved by existing orientation tools. The problem is real, the demand is clear, and language-native AI is a genuine unlock.
- Gemini's dual mode (conversational text vs. structured JSON via
responseMimeType) is powerful but requires careful orchestration to avoid breaking the conversation flow.
What's next for Mowajihi AI
- Bulletin upload: OCR-based Bac results card scanning so the AI can read grades automatically without the student typing them.
- Psychological quiz mode: A structured MBTI-style quiz as an alternative entry point for students who prefer not to chat.
- Push notifications & deadline tracking: Alert students about college registration deadlines and dossier submission windows for their recommended schools.
- Counselor dashboard: A web portal for school counselors to review aggregated (anonymized) orientation trends across their class.
- Expansion beyond Morocco: The architecture generalizes to any country with a national exam system — Algeria (Bac), Tunisia (Bac), and Senegal (BFEM/BAC) are natural next targets.
- Offline-first PWA: Many students in rural Morocco have intermittent connectivity; a service-worker-cached PWA would make Tawjihi accessible without a stable internet connection.
Log in or sign up for Devpost to join the conversation.