DZ Alumni — AI Academic Opportunity Platform
AI platform that helps students discover, match and apply to international scholarships and fellowships, with automated CV/cover-letter adaptation and intelligent scoring.
Overview
DZ Alumni is a full-stack, AI-powered platform that helps students discover, match and apply to international academic opportunities. A dual-engine scraper — Playwright for JavaScript-rendered pages, BeautifulSoup for static HTML — continuously crawls 20+ academic portals (DAAD, Chevening, Fulbright, Scholars4Dev and more), normalizing and de-duplicating results with multi-language (EN/FR/AR) parsing and an automated confidence score. Each opportunity is embedded with OpenAI and ranked against the user's profile by a 7-factor weighted scoring engine (semantic similarity, country, degree, domain, fee, deadline, funding). GPT-4 then adapts the user's CV and cover letter per opportunity — rendered to publication-quality PDF via WeasyPrint — and drafts application emails. The async FastAPI backend (SQLAlchemy, 25+ models, Alembic) follows a layered routes → services → models architecture, with Celery + Redis driving a 24/7 background scrape scheduler, JWT + Google/LinkedIn OAuth, AES encryption of sensitive data, SlowAPI rate limiting, and a community/messaging layer — all served to a Next.js 14 + React Query + Zustand frontend.
Outcomes
- Continuously scrapes 20+ academic portals with multi-language parsing and de-duplication
- 7-factor scoring engine fuses semantic embeddings with structured filters
- GPT-4 adapts CVs and cover letters per opportunity, rendered to publication-quality PDF
Key Features
- Dual-engine scraper (Playwright + BeautifulSoup) across 20+ academic portals with anti-bot measures
- Multi-language (EN/FR/AR) extraction, de-duplication and per-result confidence scoring
- Semantic matching via OpenAI embeddings (1536-dim, cosine similarity)
- 7-factor weighted relevance score (semantic, country, degree, domain, fee, deadline, funding)
- GPT-4 CV & cover-letter adaptation per opportunity → PDF (WeasyPrint)
- AI application-email generation with tone & key-argument extraction
- Deadline calendar + application pipeline (planned → submitted → accepted/rejected)
- Community forums, votes and 1-to-1 messaging (text/image/voice)
- Async FastAPI backend: 25+ models, layered services, Alembic migrations
- Celery + Redis 24/7 scheduler; JWT + Google/LinkedIn OAuth; AES-encrypted sensitive data
Role
Full-Stack & AI Engineer
Challenge
Automate the painful parts of applying to academic opportunities worldwide — discovery, relevance ranking, and per-program document tailoring.
Solution
Built a dual-engine scraper feeding an async FastAPI backend, a 7-factor scoring engine that fuses semantic embeddings with structured filters, and a GPT-4 document adapter that rewrites CVs/letters per opportunity, with regex fallbacks for resilience.