Improve developer productivity using AWS GenAI capabilities

Presentation at AWS Community Day Turkey

1. Improve developer productivity
using AWS GenAI capabilities
Matt Lewis
AWS Hero
AWS Chief Architect
IBM Consulting

2. The Rise of AI Coding Assistants
“60% of teams reported being short of engineering
resources needed to accomplish their established goals”
- Jellyfish ‘State of Engineering Management Report 2023`
“75% of enterprise software engineers will use AI coding
assistants, up from less than 10% in early 2023”
- Gartner Research

3. •Provenance of training data including AWS internal source code and
external open-source code
•Richest body of knowledge for native AWS service integrations
•Pro tier offers IP indemnity for its output
Amazon Q for Developer

4. SWE-Bench coding capability benchmark
Test a tools ability to solve real GitHub issues automatically, with verification of unit tests results

5. Amazon Q for DeveloperAmazon BedrockAmazon SageMaker
•Integrated into IDE
•No model choice for user
•Create a customisation to receive
suggestions based on team’s
internal libraries and code style
•Fully managed service offering a choice
of foundation models.
•Custom model import (preview).
•Supports customisation through
continued pre-training and fine-tuning,
with control over hyperparameters.
•Pricing tied to choice of model.
•Fully serverless offering with no
infrastructure to manage.
•Model access on demand through
Bedrock API’s.
•Supports pre-trained open-source models
and capability to train your own model.
•Supports full MLOps lifecycle, with model
governance, auditability, and automation
of machine learning workflows.
•Extensive selection of built-in algorithms
and wide array of machine learning
languages and frameworks.
•Catalog and manage model versions with
Model Registry and detect data drift.
•Advanced customisations with techniques
like model pruning and quantization.
•Allows for choice of compute including
spot instances with ability to train on AWS
Trainium accelerators and deploy on AWS
Inferentia accelerators.
Simple Complex

6. Decision Tree for Service Selection
Start
Are you using
a popular
programming
language?
Do you need a big
context window
to analyse a large
codebase?
Use Amazon
Bedrock
Use Amazon
Bedrock
Use Amazon Q
for Developer
yes
yes
no
no

7. Core Capabilities for improving DevEx
Using AWS GenAI capabilities to help a developer:
•Create code
•Ensure code is secure
•Understand an application
•Modernise an application
•Implement a new feature

8. •Code Completion
•Code Generation
•Customizations
•Command Line
Creating Code

9. •Code Completion
•Code Generation
•Customizations
•Command Line
Creating Code - Completion

10. •Code Completion
•Code Generation
•Customizations
•Command Line
Creating Code - Generation

11. •Code Completion
•Code Generation
•Customizations
•Command Line
Creating Code - Customization

12. •Code Completion
•Code Generation
•Customizations
•Command Line
Creating Code – Command Line

13. •Amazon Q can scan your codebase for security vulnerabilities and code quality
issues
•Generate a finding with a description of the issue and recommended fix
•Scans powered by security detectors informed by years of AWS and
Amazon.com security best practices
•Includes CloudFormation and Terraform
•Outperforms leading publicly benchmarkable tools on detection across most
popular programming languages.
Ensuring Code is Secure

14. Security Scan

15. Code Understanding

16. Application Visualisation

17. Code Modernisation

18. Amazon Q Developer Agent for Software Development

19. Amazon Q Developer Agent for Code Transformation

20. Thank you!
Matt Lewis
m_lewis
mattelewis