When I heard that AWS was about to release their "Certified AI Practitioner" Exam, I told myself I was going to jump on the bandwagon. There is a lot of hype around AI, Machine Learning, Foundation Mdels, LLM's. It's difficult to view through the smoke or discern who is peddling snake oil, what is easy to tell is that AI continues to make a deep impact in the Information Technology realm. Below are my 'personal' notes while studying for the Exam.
- What is GenAI
- Amazon Bedrock
- Foundation Models
- Foundation Model Evaluation
- Retrieval Augmented Generation(RAG) & Knowledge Base
- GardRails
- AI Stylist
- Prompt Engineering
- What is It
- Performance Optimization
- Techniques
- Amazon Q
- Business
- Apps
- Developer
- Kinda like Co-Pilot.
- AI and Machine Learning
- AI, ML, Depp Learning and GenAI
- Training Data
- Supervised Learning
- Tagged Input
- Unsupervised Learning
- Mix of Tagged and Untagged, mostly Untagged
- Reinforcement Learning
- Kinda like a pet, rewards good behavior, reinforces good behavior
- Model Fit, Bias, and Variance
- Inferencing
- Infer , things like in DB .. when you don't have all of the access but you can infer data points.
- Phases of a Project
- Managed AWS Services
- Comprehend
- Comprehends the written text, Natural Language Processing Exracts insight about the contents of documents.
- Translate
- Translate from languages.
- Transcribe
- Speech to Text
- Polly
- Text to Speech
- Rekognition
- Recognize things in Images and Video.
- Forecast
- Lex
- Can make Bots with this.
- Connect
- This is a Call Center type of service where you can have agents take calls, you can connect this to services like Lex and Polly.
- Personalize
- Textract
- Extract Text from documents
- Kendra
- Can take documents and data to create an enterprise search engine.
- Mechanical Turk
- This service connects real humans to do tasks for training models or double-checking AI work.
- Augmented AI
- Double Check the work of AI
- DeepRacer
- You actually race a car with a reinforcement model, you get to train your model to see what is faster at learning/getting past the finish line.
- Comprehend Medical
- Transcribe Medical
- Hardware For AI
- Tranium
- Other Chip
- SageMaker
- Complete solution for ML, can train, a model , tailor it.
