Project 12 · AIOps · Observability · GenAI

CloudPulse AI

Infrastructure Interrogation Engine

Ask your AWS infrastructure anything in plain English. CloudPulse AI fetches live CloudWatch telemetry and uses Amazon Bedrock (Claude 3) to diagnose incidents, surface anomalies, and explain root causes — in seconds, not hours.

View on GitHub See It In Action ↓

CloudPulse AI — Live Infrastructure Query

👤

Why did my Lambda function spike at 2am last night?

⚡

Fetching CloudWatch Logs for us-east-1...

Root Cause Identified: Your process-orders Lambda experienced a 3.2x invocation spike at 02:14 UTC triggered by an upstream SQS queue backlog (14,280 messages). Cold start latency averaged 847ms — 4x above baseline.

Recommendation: Enable Provisioned Concurrency for this function or increase reserved concurrency to prevent future cold start cascades during off-peak queue drains.

The Challenge

Debugging cloud infrastructure
takes hours. It shouldn't.

Engineers spend enormous amounts of time writing CloudWatch Insights queries, parsing log streams, and correlating metrics across services — just to answer a single question about what went wrong.

The Problem

Hours Lost to Log Archaeology

When an incident occurs, engineers must manually navigate CloudWatch, write Insights queries in a proprietary syntax, correlate metrics across multiple services, and interpret raw log data — a process that can take hours and requires deep AWS expertise that not every engineer on the team has.

The Solution

Ask in Plain English

CloudPulse AI accepts natural language queries, automatically fetches the relevant CloudWatch metrics and logs, and passes the live telemetry to Amazon Bedrock (Claude 3) for AI-powered root cause analysis. Any engineer — regardless of AWS expertise — can interrogate their infrastructure in seconds.

System Design

How CloudPulse AI Works

A fully serverless AIOps pipeline. Natural language in, root cause analysis out — powered by live AWS telemetry and Claude 3's reasoning capabilities.

💬

NL Query

Plain English question from engineer

→

🔗

API Gateway

Receives & routes query

→

⚡

Lambda

Fetches live CloudWatch data

→

🧠

Bedrock Claude 3

AI root cause analysis

→

📋

Diagnosis

Actionable incident report

cloudpulse/interrogate.py

# CloudPulse AI — Natural language infrastructure query
import boto3, json

cloudwatch = boto3.client('cloudwatch')
logs = boto3.client('logs')
bedrock = boto3.client('bedrock-runtime')

def interrogate(natural_language_query):
  # Step 1: Fetch live telemetry from CloudWatch
  metrics = cloudwatch.get_metric_data(
    MetricDataQueries=build_queries(natural_language_query),
    StartTime='-3h', EndTime='now'
  )

  # Step 2: Send telemetry + query to Claude 3
  response = bedrock.invoke_model(
    modelId='anthropic.claude-3-sonnet-20240229-v1:0',
    body=json.dumps({
      'messages': [{'role': 'user', 'content':
        f"Query: {natural_language_query}\n\nMetrics: {metrics}"}]
    })
  )

  # Step 3: Return AI-generated root cause analysis
  return json.loads(response['body'].read())['content'][0]['text']

Technology

AWS-Native AIOps Stack

Every component is purpose-built for cloud observability — combining live telemetry with frontier AI reasoning to deliver instant infrastructure intelligence.

Amazon Bedrock

AI inference & reasoning engine

Claude 3 Sonnet

Root cause analysis model

CloudWatch

Live metrics & log telemetry

AWS Lambda

Query processing & orchestration

API Gateway

Natural language query endpoint

Terraform

Infrastructure as Code

Business Impact

Dramatically reduces MTTR.
Empowers every engineer.

CloudPulse AI democratizes infrastructure observability — removing the AWS expertise barrier and turning hours of debugging into seconds of conversation.

⚡

Seconds to Root Cause

What previously required hours of CloudWatch query writing and log parsing is now answered in seconds. Ask "Why did my Lambda spike at 2am?" and get a precise, actionable diagnosis.

🧑‍💻

Democratized Observability

Any engineer — frontend, backend, or junior — can interrogate complex cloud infrastructure without knowing CloudWatch Insights syntax or AWS service internals.

🔍

Live Telemetry Context

Unlike static dashboards, CloudPulse AI fetches real-time metrics at query time — ensuring every diagnosis is based on current infrastructure state, not stale snapshots.