Real Life Customer Stories with AI
summary
Built an employee‑attrition classifier that prioritizes recall to avoid missed leavers, comparing scikit‑learn vs Azure ML Studio and matching a boosted decision forest by reimplementing parameters with AdaBoost, then serving the model via Flask in Docker. A Ghana project trains a Keras classifier (Inception v3) using 4 million Bing tiles, 6,000 labeled (8 people × 4 hours = 36 person‑hours) to detect small‑scale mines. Retail benchmarking uses 33,000 train / 10,000 test images and hits 95.5% top‑4 accuracy; custom translation moves from statistical to neural models with noticeable quality gains.
AI-written summary · about
chapters (8)
- 0:00Start
- 1:13About Myself
- 4:48Hr Focus
- 5:23Employee Retention
- 7:34Host and Deploy the Model
- 17:10Hyper Tuning of Hyper Parameter
- 31:40Using Pre-Built Ai Technologies
- 36:45How Long Was the Ghana Project and Person Days Needed To Deliver an Ml Model
related talks
50mBuilding Intelligent Mobile Apps using Machine Learning
35mBreaking black-box AI
02h 03mFull stack AI design and development
59mPredicting the future as a service with Azure ML and R
01h 05mAzure Machine Learning for the Developer
24mDon't want to miss out on the AI wave?
50mBuilding Intelligent Mobile Apps using Machine Learning
53mImplementing AI using Azure Machine Learning Services