Product Designer · Singularity Hub · 2024

AI Python Trainer

IDE as a textbook

Students learn Python in an IDE — an environment built for professional developers. The interface is intimidating. Errors go unexplained. Feedback is red text in the console.

Singularity Hub is a trainer that looks like an IDE but behaves like a textbook. Adaptive difficulty, a built-in AI assistant, Codewars-style challenges. Part of the Skyeng education ecosystem.

Python topics

GPT-4o assistant

visual style concepts

feedback mechanics

Level calibration

At the start, the algorithm assigns tasks across different topics: loops, arrays, strings, conditionals. Based on results, it determines the student's level and builds an individual plan. If a student struggles — the system checks prerequisites. Can't handle arrays? Start with loops first. Each topic shows mastery percentage and potential growth: "Mastered 50%. +25% if you solve this task."

Calibration: topic routing, prerequisite checks, mastery percentage per topic

Browser IDE — task, code editor, tests, AI assistant

Browser IDE: task description, code editor, tests, AI assistant in the left panel

Codewars alternative — coding challenges in trainer format

Codewars-style format: coding challenges with adaptive difficulty

AI assistant

A chat interface in the IDE's left panel. Suggest chips: "I don't understand the problem," "What's wrong with my code?", "Where do I start?" The AI explains errors using the Socratic method — asks questions, doesn't give the answer. Help has a cost: tasks solved with AI earn fewer points. Balance between support and independence. After each AI response — inline feedback: "Helped" / "Didn't help." Data feeds back into prompt improvements.

AI assistant: chat in IDE panel, question suggestions, Socratic explanation method

Submission states: Pending, Accepted, attempt history — 7+ interface states

Student dashboard

Personal analytics: 18 topics with mastery percentages, a weekly progress chart, task history with attempt counts. The student sees: 41.1% overall progress, 17 of 18 solved correctly, 5 of 18 topics studied, 3 questions asked to AI. Leaderboard ranks peers by progress. Multi-subject navigation: Python, Math, Physics, English.

Student dashboard: progress across 18 topics, task history, leaderboard

Teacher dashboard — student management, analytics, recommendations

Teacher dashboard: student analytics, content recommendations

Math module

GPT-4o generates problems and checks solutions. Graph construction: "Plot the parabola y = (x + 1)(x − 3)." Students can upload photos of handwritten solutions.

A custom math keyboard with five modes: Basic, ABC, Geometry, Inequalities, f(x). Four platform variants designed: desktop, minimal mobile, streaming, audio. Answer validation: correct, incorrect, loading states.

Math module — problem, AI chat, keyboard

Mobile math trainer — iOS prototype

Multi-platform — desktop, mobile, streaming, audio

Four platform variants: desktop, minimal mobile, streaming, audio

4 math platform variants — Desktop, Mobile Min, Stream, Audio

Math module: GPT-4o Desktop, Mobile Min, Stream and Audio — full platform set

Feedback system

Five mechanics, each context-specific:

After solving — star rating + difficulty (Easy / Normal / Hard)
After skipping — "Why are you skipping?" + AI offer: "Want to try with AI?"
Complaint — bug reports by category: Usability, Tasks, AI, Other
Monthly NPS — 3 steps: recommendation → usability → difficulty + AI usefulness
Theory — "Clear" / "Unclear" under each video

Five mechanics: task rating, skip reason, complaint, NPS, theory feedback

My role

The sole designer on a team of four developers. Designed every screen: IDE, calibration, dashboard, AI assistant, math module across four platforms. Ran UX tests, explored 9 visual directions, shipped the Material Design 3 system to production.

Singularity Hub

← Osme Singularity Words →