How to Build a Quokkabot
summary
A working quokka-identifying chatbot that accepts WhatsApp messages and emails, uses Twilio WhatsApp API and SendGrid for I/O, runs on Azure Functions, and classifies images with Azure Custom Vision (trained on roughly 286 quokka / 545 non‑quokka images). It returns percent probabilities, logs the last five results via Twilio Sync, accepts simple error reports, and can be retrained; demos mostly succeed but the model sometimes confuses similar animals (wombats, squirrels) and was gradually improved by adding more training images.
AI-written summary · about
chapters (11)
- 0:00Introduction and background
- 2:52Project architecture
- 6:42Building the messaging bot
- 10:05Image handling and logic
- 15:05Integrating machine learning
- 21:54Processing classification results
- 27:11Improving user experience
- 30:57Adding email communication
- 37:20Real-time sync and display
- 41:15Future work and use cases
- 45:35Q&A and conclusion
related talks
55mHow to Build a Quokkabot
20mHow to Build a Quokkabot
35mChatbots: The hype, challenges and opportunities
39mUsing WhatsApp as a Command Line ( Breaking out of the walled Garden)
10mThe future is here, and it started 5000 years ago
55mEnterprise Chatbots (using messaging and AI to build novel user interfaces for work)
33mBuilding the next-generation Conversational AI with Python and Deep Learning
35mAI JavaScript Rocks