Live Coding A Machine Learning Model from Scratch

on YouTube ↗39m

" summary

Builds a multi-label text classifier for five Stack Overflow tags using TensorFlow tf.keras and a 400-word bag of words vocabulary extracted from BigQuery Stack Overflow data (uses ~150k train / 37k test rows). Code uses Pandas, scikit‑learn and Colab, replaces explicit tag names with “avocado” during preprocessing, saves tokenizer with pickle and model as H5, reaches about 96% accuracy on held-out test data, explains predictions with SHAP, and deploys inference with Cloud AI Platform + Cloud Functions (Chrome extension demo falls back to a video).

AI-written summary · about

" chapters (16)

1. 0:00Introduction
2. 4:32Getting Data
3. 11:29Demo
4. 11:59Collab Setup
5. 14:05Encoding Data
6. 15:14Train Test Split
7. 16:16Tokenizer
8. 18:56Create Model
9. 22:41Train
10. 23:38Evaluation
11. 25:47Custom Code
12. 27:00Deploy
13. 30:14Predict
14. 32:31Summary Plot
15. 34:05Pandas Question
16. 35:45Live Demo

" related talks

• 38mReal Life Customer Stories with AIRina Ahmed · WeAreDevelopers AI Congress Vienna 2018
• 35mIt's Alive! Machine Learning Writes Your Code!Dominic Elm · AngularConnect 2019
• 30mKick-Start your Understanding of Machine Learning with PythonAndreas Buhlmeier · ML Conference 2018 Spring
• 37mMachine Learning on Source CodeFrancesc Campoy Flores · Scale By The Bay 2018
• 43mGetting started with AI for freeDan Barker · REFACTR.TECH 2019
• 01h 38mNavigating the World of NLP and Text AnalyticsIsrael Odeajo · DSC MENA 2024
• 42mDevOps & Microservices: “Delivering Machine Learning models early and often.”Kristaps Bebris · Riga Dev Days 2019
• 02h 03mFull stack AI design and developmentKrishnav Dave · SkillUp 2021