Path certificates — preview
These are all the certificates earned on the AI Engineer path: a capability certificate per stage, and the path certificate at the end. Each one names the capability you demonstrated — not the hours you watched.
After this stage you can explain how an LLM works and judge where AI fits in a product — and where it doesn't.
After this stage you can ship a real streaming, error-handled, cost-budgeted LLM feature.
After this stage you can build a RAG Assistant that answers from real documents, and diagnose retrieval and hallucination problems.
After this stage you can build a system that uses tools safely, with structured outputs and human escalation.
After this stage you can build an agent that plans, uses tools, and completes multi-step tasks reliably.
After the Capstone you have a real, published AI system with evidence and a score — proof you're an AI engineer.
Built and shipped a real AI system: LLM apps, RAG, tools, agents, and evaluation — with evidence and a score.
Explain how an LLM works, and judge where AI belongs in a product — and where it doesn't.
Pick the right model, read your bill in tokens, and manage the context window before it fills.
Write structured prompts, test them systematically, and recognize prompt injection on sight.
Ship a working feature with streaming, secured keys, and retries that don't storm.
Build an assistant that answers from your documents with citations, and diagnose weak retrieval.
Finished the path with working AI features you built yourself — from a first API call to RAG that answers from your own documents.
Write a tool schema the model understands, and control when it calls a tool — and when it shouldn't.
Get valid JSON back every time, and handle schema drift before it takes production down.
Build a loop that plans, acts, and stops — with explicit stop conditions and a budgeted context.
Give an agent real knowledge: decide when to search, decompose the question, and cite what it found.
Connect an agent to external tools over MCP with least privilege, and close the injection door in tool descriptions.
Measure a non-deterministic system against a golden set, cap its cost, and stop an agent from looping forever.
Design, build, and debug tool-using agentic systems — with guardrails, a cost ceiling, and real evaluation.
Stages 6 and 7 (Reliability and Production) are assessed within the Agents and Capstone stages — they don't carry a separate certificate.