// how it works
How Safua Works
Safua works like a medical residency for AI engineers. Three phases that take you from fundamentals to verified, production-grade capability. No lectures. No certificates. Just proof.
// phase 01 — foundation
Prove Your Baseline
Before you touch any school, prove you can write code. Six foundational domains calibrate your level and unlock the real work.
Python
Git
APIs
SQL
Docker
AI Fundamentals
Diagnostic Calibration
Each domain starts with a diagnostic. The system calibrates your level and skips what you already know. No wasted time on basics you have already mastered.
Foundation Faculty
Kaia
Lead Instructor
Nadia
Teaching Assistant — RAG & Retrieval
Hiroshi
Teaching Assistant — Fine-Tuning
Three speeds. One standard.
Foundation adapts to your experience level — but everyone passes the same checkpoint.
Standard Track
4–6 weeks
Structured lessons, guided exercises, and mentor support. Ideal if you are building these skills for the first time.
Accelerated Track
1–3 weeks
Diagnostic places you ahead. You skip what you already know and focus on gaps. For engineers with partial coverage.
Proving Track
2–5 days
Straight to the checkpoint. Prove mastery across all six domains in one sprint. For experienced engineers who just need the credential.
The proof profile shows “Foundation Complete” — not which track you took. The credential is the same. The speed is yours.
// phase 02 — build
Work Like an Engineer
You are assigned to a virtual company. You receive real tickets. You ship code inside a production-like environment. This is where the learning happens.
NovaMind AI
Agent orchestration, generative AI, retrieval quality
Fast iteration and hallucination control at the bleeding edge of AI.
DataForge Labs
High-throughput data engineering, real-time pipelines, observability
Scale and reliability under production pressure.
VisionArc
Deep learning, computer vision, edge inference
Performance under hardware and latency constraints.
Sentient Health
Healthcare AI, explainability, compliance
Building AI that meets strict regulatory requirements (HIPAA, audit trails).
Company Tickets
Real work. Real codebase.
You receive tickets scoped to a virtual company. Requirements, constraints, and deadlines mirror actual engineering work.
Execution Cockpit
Cloud IDE. No setup friction.
A full development environment with terminal, file explorer, and live preview. Write, run, and test code without leaving the browser.
AI Code Review
5-dimension scoring.
Every submission is evaluated across correctness, code quality, architecture, edge cases, and communication. No multiple choice. No shortcuts.
Build Phase Faculty
Dimitri Volkov
VP of Engineering, NovaMind AI
“This isn’t a coding exercise. The retrieval team is blocked until your pipeline ships.”
Rafael Mendes
Senior Engineer — RAG & Retrieval, NovaMind AI
“Your chunking strategy tells me everything about how you think about this problem.”
// phase 03 — prove
Build Your Verified Profile
Every line of code you write is evaluated. Five scoring dimensions build a Confidence Score that proves exactly what you can do.
Correctness
Code Quality
Architecture
Edge Cases
Communication
Confidence Score
A 0.00 to 1.00 rating of how certain the review engine is about your evaluation across all five scoring dimensions. Higher consistency means higher confidence.
Public Profile
Your Proof-of-Competency profile is a shareable link that shows verified skills, scores, and project history. No self-reported claims. Just data.
Prove Phase Faculty
Dr. Yuki Tanaka
CTO, NovaMind AI
“Would you deploy this to production with your name on it?”
Jordan Hayes
Career Director, Safua Career Office
“Your resume says ML Engineer. Let me show you what the market says about your actual skill distribution.”
Ready to start?
Begin with Foundation. Prove your baseline. Then build real systems.
Start Building