10 Ways to use ML with GCP

https://www.youtube.com/watch?v=oQMgqMRR-io

TLDR: basically an advertisement for VertextAI

1. Prototype to classify images, specifically for identifying flower species based some photos I have.

1-1 Teachable Machine - browser base tool classifiers - capture training data directly. 
https://teachablemachine.withgoogle.com/train/image

2. AutoML Vertax AI

Custom enterprise levle - text, image, tabular data.

 

3. ML powered capabilities without needing to build or deploy any models. e.g. Image labelling, sentiment analysis, text classification.

- Pre-trained ML APIs. - Single API call. Vision API, object eprson detect from video, landmark detection etc. 

 

4. Generate images and text
- Generative AI Studio. Vision AI Generation. Tune foundation model.

 

5. Search, discover 

Vertex AI Model Garden - single env to interact with a variety of model types. Moels will span across modalities. Deploy to endpoints with ease.

Code generation & Completion, Image Generation, Universal Speech, Embeddings.

 

6. Reinforcement Learning from Human FEedback - RLHF

7. BigQuery - to train my own ML model.

BigQuery ML.

Using SQL - classification model, prediction, ml.dot predict query.

I can access to image data by creating an object table - GCS. 

 

8. Playground for ML, Python

Colab - 

 

9. Integration with Cloud.

Vertex AI workbench - pre-installed with GPU. 

 

10. Custom model at scale.

Vertex AI custom training. 

 

1. Embeddings - search data

Vertex AI maching Engine.

vector database for embeddings. 2 million random images - search images from this yellow car- maching engine finding the nearest embeddings. return the results in milleseconds.