Doing Machine Learning Without a Data Scientist
summary
Describes how to democratize machine learning by avoiding full model training and instead reusing pretrained APIs and simple heuristics. Gives a failed example where training a rock–paper–scissors classifier on 200 images only reached ~33% accuracy, and contrasts that with a simpler symbolic gesture detector (horizontal vs vertical). Presents a concrete project that polls Twitter timelines and uses Google Vision API and Natural Language API to score entities and sentiment, then compares pattern distances to detect "filter bubble" similarity.
AI-written summary · about
chapters (8)
- 0:00Machine learning in business
- 2:50Quick wins first
- 4:15Failed image recognition
- 6:43Using ML APIs
- 8:55NLP bubble analysis
- 13:29Scoring tweets with NLTK
- 16:00Value without data science
- 17:44Company intro
related talks
20mDoing Machine Learning Without a Data Scientist
30mMachine Learning: Beyond the Tutorial
06mInterview
54mMachine Learning explained with Examples
32mThe Miseducation of This Machine
47mBuilding smarter solutions without machine learning expertise
42mUse Machine Learning in your code, without being a ML expert
40mMachines are Learning