AI Instructor – RAG

YPAI; United States; Remote; USD 36,000 - 136,000; Full time

Job summary

YPAI Learn offers live online classes for developers, teaching retrieval-augmented generation with citations. Classes are 90 minutes each, conducted in English and delivered remotely using a class kit that includes slides, a plan, and demos.

Instructors guide learners through chunking documents, embeddings, and hybrid search methods, while evaluating with tests and source-grounded answers. Compensation is per class and paid biweekly via PayPal, Wise or bank transfer.

Qualifications

• Experience teaching, mentoring engineers, running hands-on labs or giving technical talks.
• Fluent spoken English is required to teach in English.

Responsibilities

• Teach the live class twice: “Chunk, embed and search a document set to answer with citations” (90 minutes) and “Score retrieval and grounding on a test set, then fix what fails” (90 minutes).
• Demonstrate how chunk size, overlap and metadata affect retrieval results and how hybrid keyword and vector search with a reranker improves accuracy.
• Guide learners to build test sets and score recall at k, faithfulness and answer relevance after each change.
• Assist learners when issues occur (near-duplicates, empty pages, or answers beyond sources) by inspecting retrieved chunks.

Skills

Teaching experience

Mentoring engineers

Hands-on labs

Technical talks

Fluent English

Tools

Vector database

Job description

Teach live online classes for developers on retrieval-augmented generation: chunking documents, embedding and indexing them, retrieving the right passages and grounding answers in cited sources. You teach in English using a ready class kit, remotely, and you are paid for each class you teach.

What the work is for

YPAI Learn gives developers live classes in building AI that answers from their own documents. Your classes take them from raw PDFs and web pages split into chunks to a search that returns cited passages, and teach them to measure whether the right passages came back and whether each answer stays within them.

The work

• Teach the live class “Chunk, embed and search a document set to answer with citations”, 90 minutes, from the class kit.
• Teach the live class “Score retrieval and grounding on a test set, then fix what fails”, 90 minutes, from the class kit.
• Show live how chunk size, overlap and metadata change what a query retrieves, and how hybrid keyword and vector search with a reranker lifts the right passage to the top.
• Walk learners through building a test set of questions with known source passages, then scoring recall at k, faithfulness and answer relevance after each change.
• Help learners when search returns near-duplicates, a table or scanned page comes back empty, or the model answers beyond its sources, by inspecting the retrieved chunks first.

What to expect

• All classes are live, online and in English, open to learners in any country. Other languages you can teach in go in a separate question on the application.
• Each kit holds a minute‑by‑minute plan, slides and speaker notes, a demo of retrieval over a sample document set, and a backup recording of that demo.
• You teach remotely on your own computer, with a steady connection, a camera and a microphone, and a development setup that runs the demo notebook or service.
• Each class has an instructor test, and you pass it before your first time teaching that class.
• Before a teaching assignment starts, you sign an instructor agreement. It covers pay, confidentiality and how class material may be used.
• Each invitation states the class times and the price learners pay, and you see both before you accept.

What you bring

• A retrieval system you built over real documents, with its repository, evaluation results or a write‑up you can walk through.
• Experience teaching, mentoring engineers, running hands‑on labs or giving technical talks.
• Patience with developers who expect a vector database to fix answer quality on its own.

You need

• Hands‑on work building retrieval‑augmented generation in real projects: parsing, chunking, embeddings, vector or hybrid search and reranking.
• Experience measuring answer quality with test sets, retrieval metrics and checks that answers stay grounded in their sources.
• Fluent spoken English, the language the classes are taught in.
• 18 years or older.

Good to know

• If outside work needs sign-off under your employment contract, arrange the sign‑off before accepting any assignment.
• Demand for the classes sets how much teaching there is. The expected time commitment is written in each invitation.
• Applying is free of charge; you need not buy or take a YPAI Learn course.

Pay

$26–$99/hr. Rate shown is for the United States. This role is open in 4 countries, with local rates; pick your country on YPAI.

YPAI pays every two weeks for approved work via PayPal, Wise or bank transfer. Work is by invitation after your profile is reviewed.