Atlas Engine - AI-Powered Sales Acceleration Platform

Deploy an AI-powered sales automation system to AWS in 30 minutes.

🚀 One-Click Deployment

Deploy directly to your AWS account:

Note: This button uses the manually uploaded template. Lambda code must be deployed separately using ./deploy.sh

About Atlas Engine

What if you could engage a new web lead not in hours, but in seconds?

The Atlas Engine is a serverless, AI-powered B2B sales accelerator built on AWS. It's a high-impact, interactive demo that solves the "immediate contact" problem by turning a simple web chat into a real-time, AI-driven, outbound phone call. It's a "silent salesperson" that demonstrates a sophisticated, event-driven architecture in real-time.

The Journey (Web Chat to Phone Call in < 60 Seconds) goes as follows:

1. A user interacts with an Amazon Lex chatbot on the web.

2. Their info triggers an AWS Step Functions workflow, which creates a lead in Salesforce.

3. Amazon Bedrock (Claude 3.5 Sonnet) dynamically generates a contextual greeting based on the user's chat.

4. Amazon Connect is triggered, placing an immediate outbound, AI-powered voice call to the user.

5. The user talks to the AI-driven IVR, which can answer questions live (using Bedrock) or handle requests.

6. After the call, Amazon Transcribe and Bedrock create a summary, and the Step Function resumes to post it to the Salesforce lead record.

Prerequisites for One-Click Deploy

Before clicking the deploy button, ensure you have:

1. AWS Account with admin permissions
2. Bedrock Model Access - Enable Claude 3.5 Sonnet in Bedrock Console
3. Salesforce Connected App - Create credentials following SALESFORCE_INTEGRATION.md
4. AWS Secrets Manager Secret - Store Salesforce credentials:

   aws secretsmanager create-secret \
       --name AtlasEngine/SalesforceCreds \
       --secret-string '{"username":"your-email@company.com","client_id":"YOUR_CLIENT_ID","private_key":"-----BEGIN PRIVATE KEY-----\nYOUR_KEY\n-----END PRIVATE KEY-----"}'

Expected Costs

Environment	Monthly Cost	Use Case
Development	~$24/month	Testing, 1K conversations
Staging	~$214/month	QA, 10K conversations
Production	~$1,138/month	Live, 50K conversations

See COST_ESTIMATION.md for detailed breakdown and optimization strategies.

Post-Deployment Steps

After CloudFormation completes:

1. Configure Amazon Lex Bot

   • Go to Amazon Lex Console
   • Create a new bot named "AtlasEngineBot"
   • Add the Lambda fulfillment ARN from CloudFormation Outputs
   • Build and publish the bot

2. Optional: Set Up Amazon Connect (for voice calls)

   • Go to Amazon Connect Console
   • Create or use existing instance
   • Update stack parameters with ConnectInstanceId and SourcePhoneNumber
   • Configure contact flow to use the deployed Lambda functions

3. Test the System

   # Test Step Functions workflow
   aws stepfunctions start-execution \
       --state-machine-arn <WORKFLOW_ARN_FROM_OUTPUTS> \
       --input '{"firstName":"John","lastName":"Doe","phone":"+15555551234"}'

---

👉 CONFUSED? Pick your guide:

• Never done this before? → START_HERE.md
• Just want commands? → CHEAT_SHEET.txt
• Want full details? → Keep reading below

🚀 How to Deploy (Step-by-Step)

Step 1: Open Terminal

1. Open Terminal on your Mac
2. Navigate to this folder:

   cd /Users/johndoe/Downloads/aws_org_export

Step 2: Configure AWS

The scripts will use your AWS CLI configuration. To set region:

# Check current region
aws configure get region

# Change region (if needed)
aws configure set region us-west-2

Note: You don't need to manually add account ID - AWS automatically detects it from your credentials.

Step 3: Set Up Salesforce Credentials

Create a file with your Salesforce credentials:

cat > salesforce-creds.json << 'EOF'
{
  "username": "your-email@company.com",
  "client_id": "YOUR_SALESFORCE_CLIENT_ID",
  "private_key": "-----BEGIN PRIVATE KEY-----\nYOUR_KEY_HERE\n-----END PRIVATE KEY-----"
}
EOF

Step 4: Run Deployment

# Make scripts executable (one-time setup)
chmod +x *.sh

# Run deployment
./deploy.sh

The script will:

• Ask you questions (environment, region, etc.)
• Build everything automatically
• Deploy to AWS
• Show you what to do next

That's it! Takes 15-20 minutes.

Step 5: Test (Optional)

./test-deployment.sh dev

Common Questions

Q: Do I need to edit any files with my account ID? A: No! The scripts automatically detect your AWS account from your credentials.

Q: How do I change the region? A: Either set it before deployment (aws configure set region us-west-2) or the script will ask you.

Q: What if I don't have Salesforce credentials yet? A: The script creates a placeholder secret. You can update it later.

Q: How much will this cost? A: Development environment: ~$24/month. See COST_ESTIMATION.md.

Q: How do I remove everything? A: Run ./cleanup.sh dev

📋 What's Included

Core Components

• 8 Lambda Functions: CreateLead, GenerateScenario, InvokeCall, LexFulfillment, UpdateLead, StartTranscription, SummarizeAndResume
• 2 Lambda Layers: Python libraries (requests, phonenumbers) and Salesforce libraries (simple-salesforce, PyJWT)
• 2 DynamoDB Tables: Interactions storage and task token management
• 1 Step Functions Workflow: Orchestrates lead creation → scenario generation → outbound call → lead update
• Amazon Bedrock Integration: Claude 3.5 Sonnet for AI-powered conversations
• Salesforce Integration: JWT authentication for CRM sync
• Amazon Connect: Outbound voice contact capability

Deployment Tools

• deploy.sh: Automated deployment with validation
• cleanup.sh: Complete resource removal
• validate.sh: Pre-deployment checks
• test-deployment.sh: Post-deployment validation
• build-layers.sh: Lambda layer builder

Documentation

• QUICK_START.md: 30-minute deployment guide
• DEPLOYMENT_GUIDE.md: Comprehensive deployment documentation
• COST_ESTIMATION.md: Detailed cost breakdown and optimization
• DEPLOYMENT_ARCHITECTURE.md: System architecture and dependencies
• SECURITY_AUDIT_REPORT.md: Security findings and remediation

Templates

• sam-template/template.yaml: Main SAM template
• quick-start.yaml: One-click CloudFormation deployment
• parameters-example.json: Configuration examples

💰 Cost Estimate

Environment	Monthly Cost	Use Case
Development	~$24	Testing, 1K conversations
Staging	~$214	QA, 10K conversations
Production	~$1,138	Live, 50K conversations

See COST_ESTIMATION.md for detailed breakdown.

🏗️ Architecture

User → Amazon Lex → Lambda (Fulfillment)
                         ↓
                   Step Functions
                         ↓
        ┌────────────────┼────────────────┐
        ↓                ↓                ↓
   Salesforce      Amazon Bedrock   Amazon Connect
   (CRM Sync)      (AI Scenarios)   (Voice Calls)
        ↓                ↓                ↓
        └────────────────┼────────────────┘
                         ↓
                    DynamoDB
                  (Interactions)

📦 Project Structure

aws_org_export/
├── deploy.sh                    # Main deployment script
├── cleanup.sh                   # Resource cleanup
├── validate.sh                  # Pre-deployment validation
├── test-deployment.sh           # Post-deployment tests
├── build-layers.sh              # Lambda layer builder
├── requirements.txt             # Python dependencies
├── parameters-example.json      # Configuration template
├── quick-start.yaml             # One-click CloudFormation
├── QUICK_START.md              # 30-min deployment guide
├── DEPLOYMENT_GUIDE.md         # Full documentation
├── COST_ESTIMATION.md          # Cost analysis
├── lambda/                     # Lambda function code
│   ├── CreateLeadHandler_code/
│   ├── GenerateDynamicScenarioHandler_code/
│   ├── InvokeOutboundCallHandler_code/
│   ├── LexFulfillmentHandler_code/
│   ├── UpdateLeadHandler_code/
│   ├── StartTranscriptionHandler_code/
│   └── SummarizeAndResumeHandler_code/
├── sam-template/
│   ├── template.yaml           # SAM template
│   └── samconfig.toml          # SAM configuration
└── stepfunctions/
    └── workflow-definition.asl.json  # Step Functions definition

🔧 Prerequisites

• AWS CLI 2.x
• SAM CLI 1.100+
• Python 3.13+
• AWS account with:
  • Admin IAM permissions
  • Bedrock model access (Claude 3.5 Sonnet)
  • Salesforce Connected App credentials

📖 Deployment Options

Option 1: Automated Script (Recommended)

./deploy.sh

Option 2: Manual SAM Deployment

./build-layers.sh
sam build --parallel
sam deploy --guided

Option 3: One-Click CloudFormation

aws cloudformation create-stack \
    --stack-name AtlasEngine-Infrastructure \
    --template-body file://quick-start.yaml \
    --parameters file://parameters.json \
    --capabilities CAPABILITY_IAM

🧪 Testing

# Validate environment
./validate.sh

# Deploy to dev
./deploy.sh

# Run tests
./test-deployment.sh dev

# Test Step Functions
aws stepfunctions start-execution \
    --state-machine-arn <WORKFLOW_ARN> \
    --input '{"firstName":"John","lastName":"Doe","phone":"+15555551234"}'

🔐 Security

• Secrets stored in AWS Secrets Manager
• IAM least-privilege policies
• Encryption at rest for DynamoDB
• VPC deployment supported
• No hardcoded credentials

See SECURITY_AUDIT_REPORT.md for audit results.

🎯 Customization

Change Bedrock Model

Edit sam-template/template.yaml:

BedrockModelId:
  Default: anthropic.claude-3-opus-20240229-v1:0

Add Custom Lambda

MyCustomHandler:
  Type: AWS::Serverless::Function
  Properties:
    CodeUri: ../lambda/MyCustomHandler_code/
    Handler: lambda_function.lambda_handler

Multi-Environment

# Deploy to staging
ENVIRONMENT=staging ./deploy.sh

# Deploy to production
ENVIRONMENT=prod ./deploy.sh

🗑️ Cleanup

# Remove all resources
./cleanup.sh dev

# Or manually
aws cloudformation delete-stack --stack-name AtlasEngine-dev

📊 Monitoring

# View Lambda logs
sam logs --stack-name AtlasEngine-dev --tail

# View Step Functions executions
aws stepfunctions list-executions \
    --state-machine-arn <WORKFLOW_ARN>

# View DynamoDB metrics
aws cloudwatch get-metric-statistics \
    --namespace AWS/DynamoDB \
    --metric-name ConsumedReadCapacityUnits \
    --dimensions Name=TableName,Value=AtlasEngineInteractions-dev

🐛 Troubleshooting

Deployment fails

# Check SAM logs
sam logs --stack-name AtlasEngine-dev

# Validate template
sam validate

# Clean and rebuild
rm -rf .aws-sam && sam build

Lambda timeout

Increase timeout in template.yaml:

Timeout: 300
MemorySize: 1024

Bedrock access denied

Enable model access in AWS Console: Bedrock → Model access → Manage model access

See DEPLOYMENT_GUIDE.md for more troubleshooting.

📚 Documentation

• QUICK_START.md - 30-minute deployment
• DEPLOYMENT_GUIDE.md - Complete guide
• COST_ESTIMATION.md - Cost analysis
• DEPLOYMENT_ARCHITECTURE.md - Architecture
• SECURITY_AUDIT_REPORT.md - Security audit

🤝 Contributing

1. Fork the repository
2. Create feature branch
3. Test changes with ./validate.sh
4. Submit pull request

📄 License

MIT License - See LICENSE file for details

🆘 Support

• AWS Documentation: https://docs.aws.amazon.com/
• Bedrock Docs: https://docs.aws.amazon.com/bedrock/
• SAM Docs: https://docs.aws.amazon.com/serverless-application-model/
• Issues: Create GitHub issue

🎉 Success Metrics

After deployment, you should see:

• ✅ 8 Lambda functions deployed
• ✅ 2 DynamoDB tables created
• ✅ 1 Step Functions workflow active
• ✅ Salesforce integration working
• ✅ Bedrock models accessible
• ✅ All tests passing

Ready to deploy? Run ./deploy.sh now!