DevOps & Microservices: “Delivering Machine Learning models early and often.”
summary
Prototype that grades free-text learning quiz answers uses pre-trained word embeddings (Google/Stanford/Facebook) with Python Gensim, a tiny Flask + scikit-learn API and a human-in-the-loop labeling UI. Thirty-five lines bootstraps the service, models reach ~75% baseline on ~3,000 examples, and adding a confidence threshold yields ≈90% precision when model is confident. Deployed as a lightweight random forest microservice (Gunicorn/HAProxy, Redis optional); good enough for low-stakes instant feedback, not state-of-the-art.
AI-written summary · about
chapters (9)
- 0:00Introduction and context
- 3:53Problem with manual grading
- 11:26Building the first model
- 14:20Feedback loop and data improvement
- 20:09Model refinement and precision
- 25:11Microservices for deployment
- 28:07Integration and real-time grading
- 32:16Key takeaways and future work
- 41:25Q&A
related talks
30mProductionizing an unsupervised machine learning model to...
29mBuilding Machine Learning demos with Python
31mMachine Learning-Enabled Chatbots
40mDeploying Your Own Language Models: A Use Case Demo with Open Source...
37mDetecting offensive messages using Deep Learning: A micro-service based approach
33mHow you really get your data science models into production the cool way!
50mCreating a scalable (...) with Elixir, Python & Tensorflow
38mBuilding robust machine learning systems