SkillMateAI sits silently inside every teaching session — analysing the conversation, scoring understanding, and generating personalised quizzes from what was actually taught.
Free during the hackathon · No credit card · Open source on GitHub
Understanding
Weak topics
Generated
in ~3 seconds, on-device-class compute
3
distinct Gemma 4 calls
~3s
transcript → quiz
5
MCQs per session
0
data leaves your model
Three steps. Zero friction. Powered end-to-end by Gemma 4 running on Google AI infrastructure.
Run your usual Zoom or Meet session. Upload the chat transcript when you're done — no plugins, no recording.
The model scores understanding 0–10, surfaces weak topics, and writes a session summary in plain English.
5 MCQs are generated from the actual lesson content — not generic questions. Progress is tracked over time.
Gemma 4 reads the full chat transcript, scores understanding 0–10, and surfaces exactly which sub-topics need reinforcement.
5 MCQs auto-generated from each session, hitting the exact concepts the student tripped on — not a generic textbook quiz.
Paste any notes, slide content, or textbook excerpt — Gemma 4 generates a personalised practice set instantly.
Visualise quiz performance across every topic. Gemma 4 writes personalised feedback once enough data is in.
Students find teachers by subject, watch teaching samples, and request sessions — all without leaving the app.
Every quiz question is grounded in the actual transcript. No hallucinated facts, no off-topic drift — the model cites what was taught.
Gemma 4's open weights mean SkillMateAI works in a classroom with spotty internet, a tutoring centre that can't share student data, or a phone in a student's pocket. Same model, same quality, no compromise.
Model
gemma-4-27b-it
analyse_session()
transcript → score, weak topics, summary, 5 MCQs
generate_self_study_questions()
pasted notes → topic + 5 MCQs
generate_progress_feedback()
attempts → personalised paragraph
Sign up in 30 seconds. Run your first AI-analysed session today.