Improve developer productivity
using AWS GenAI capabilities
Matt Lewis
AWS Hero
AWS Chief Architect
IBM Consulting
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
• 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
SWE-Bench coding capability benchmark
Test a tools ability to solve real GitHub issues automatically, with verification of unit tests results
Amazon Q for Developer Amazon Bedrock Amazon 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
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
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
• Code Completion
• Code Generation
• Customizations
• Command Line
Creating Code
• Code Completion
• Code Generation
• Customizations
• Command Line
Creating Code - Completion
• Code Completion
• Code Generation
• Customizations
• Command Line
Creating Code - Generation
• Code Completion
• Code Generation
• Customizations
• Command Line
Creating Code - Customization
• Code Completion
• Code Generation
• Customizations
• Command Line
Creating Code – Command Line
• 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
Security Scan
Code Understanding
Application Visualisation
Code Modernisation
Amazon Q Developer Agent for Software Development
Amazon Q Developer Agent for Code Transformation
Thank you!
Matt Lewis
m_lewis
mattelewis

Improve developer productivity using AWS GenAI capabilities

  • 1.
    Improve developer productivity usingAWS GenAI capabilities Matt Lewis AWS Hero AWS Chief Architect IBM Consulting
  • 2.
    The Rise ofAI 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 oftraining 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 capabilitybenchmark Test a tools ability to solve real GitHub issues automatically, with verification of unit tests results
  • 5.
    Amazon Q forDeveloper Amazon Bedrock Amazon 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 forService 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 forimproving 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 Qcan 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.
  • 15.
  • 16.
  • 17.
  • 18.
    Amazon Q DeveloperAgent for Software Development
  • 19.
    Amazon Q DeveloperAgent for Code Transformation
  • 20.