Skip to main content

⬅️ Back to Project Overview

Tooling & Libraries Applied to Generative AI Implementation Lifecycle for ShieldCraft

Key tools and libraries, and how they apply to the various steps in the Generative AI implementation lifecycle for ShieldCraft.

Tools & Libraries​

  • LangChain: Orchestration and chaining of LLMs and tools
  • Bedrock: AWS managed GenAI service
  • SageMaker: Model training, deployment, and monitoring
  • MLflow: Experiment tracking and model registry
  • Poetry: Dependency management and packaging
  • Pandas, PyArrow: Data preparation and transformation
  • FastAPI: API development and integration
  • nbval, pytest: Notebook and code testing
  • ruff, black, mypy: Linting and static analysis

Application to Generative AI Implementation Lifecycle Steps​

1. Discovery & Use Case Definition​

  • Use LangChain and Bedrock to prototype and validate use cases.
  • Document requirements and success criteria.

2. Data Preparation & Retrieval Strategy​

  • Use Pandas, PyArrow for data cleaning and transformation.
  • Integrate with cloud storage and data lakes.

3. Model Selection, Prompt Engineering & Initial Prototyping​

4. Application Integration & Orchestration​

5. Evaluation, Testing & Refinement​

  • Use nbval and pytest for testing notebooks and code.
  • Track experiments and results with MLflow.

6. Deployment, MLOps & Monitoring​

  • Deploy models with SageMaker and MLflow.
  • Monitor performance and drift.
  • Manage dependencies and packaging with Poetry.