## AI Tutor

demo

Your personal AI coach — adaptive feedback on every prompt, every scenario, every eval

active session

RAG retrieval — chunking strategies

adaptive

Picked because your retrieval recall@5 dropped 12% last week.

Conversation

3

turns

Avg session: 18 turns

Tutor rating

4.9

/ 5

Across 1,247 learners

Drills unlocked

4/ 5

unlock more

Next drill: write the eval

Live session — RAG retrieval

The tutor watches every attempt and adapts its next question to your weakest sub-skill.

Today's focus: RAG retrieval. You'll write a chunking strategy for a noisy 200-page support corpus. Ready?

Lesson opener

You

I'd split by section headings, 800 tokens per chunk, 100-token overlap.

Reasonable starting point. Two issues: (1) section headings vary wildly in this corpus (some sections are 50 tokens, some 4000) — fixed boundaries break long sections. (2) 100-token overlap is fine for prose but loses structure on tables. What would you change?

Adaptive critique — caught fixed-boundary assumption

SendThe tutor remembers your last 30 days of practice.

Tutor's notes on you

What the agent has learned about your gaps. Updates as you reply.

Strength

Solid mental model on prompt structure and few-shot framing.

Gap

Defaults to fixed-boundary chunking. Hasn't tried semantic splitters yet.

Plan

Today: chunking. Thurs: retrieval eval. Mon: end-to-end RAG scenario.

Invalid domain for site key.