04 · AI SystemsTrack · Live
Agentic, grounded, shipped.
Multi-agent architectures and RAG pipelines in production.
CVDesigner
End-to-end agentic platform that analyzes job postings, researches target companies, and generates ATS-optimized CVs tailored to each role — while staying faithful to the candidate's real experience.
01
Four specialized agents
job-analysis (extracts requirements and weights), research (gathers company context via web search), generation (rewrites CV sections with structured output), and scoring (evaluates fit via custom evaluation criteria).
02
RAG pipeline on pgvector
Candidate profiles and job descriptions are embedded into pgvector for semantic retrieval, with custom skill definitions, shared context windows, and function calling across agents.
03
Quantitative fit scoring
Produces profile–role alignment percentage, key strengths to emphasize, and skill gaps to address — evidence-backed, not vibes.
- Python
- Agno
- Pydantic
- Gemini API
- pgvector
- PostgreSQL
- Next.js
- Gemma 4
- Qwen 3.6
AI Tutor
AI-powered tutoring platform that ingests curriculum documents (starting with the Moroccan national program) and generates personalized, structured courses adapted to each learner's level and pace.
01
Curriculum-grounded RAG
Official curriculum PDFs and textbooks are chunked, embedded into pgvector, and retrieved contextually to ground every generated lesson in verified pedagogical content.
02
Adaptive generation
LLM-driven content produces structured lessons, interactive quizzes, step-by-step explanations, and adaptive difficulty scaling based on learner performance signals.
03
Curriculum-agnostic
Built to scale beyond Morocco to any national program or custom course material — same pipeline, different source corpus.
- Next.js
- Python
- FastAPI
- Agno
- Pydantic
- Gemini API
- pgvector
- Supabase
- Gemma 4
- Qwen 3.6