Sanjeev KumarAI Cloud • DevOps • Wealth
Portfolio

System Engineering Architectures

Detailed case studies of infrastructure layouts, continuous automation pipelines, and Generative AI modules deployed in production.

devops
KubernetesAWS (EKS)HelmKarpenterTerraform

Production Kubernetes Cluster Setup

Problem Statement:

An e-commerce business was suffering from periodic service crashes during traffic spikes due to manual server scaling limits.

Proposed Solution:

Designed and deployed a production-grade Amazon EKS (Elastic Kubernetes Service) cluster with cluster autoscaler, Karpenter for efficient node provisioning, and Ingress controllers.

View GitHub RepoLive Architecture

Deployment Topology

[Traffic Request] -> [AWS Route 53] -> [Application Load Balancer] | +-------------------------+-------------------------+ | (Ingress Routes) | v v [EKS Node: Pod - Frontend] [EKS Node: Pod - API] (HPA Auto-scaling) (Spot Instances) | | +-------------------------+-------------------------+ | v [Amazon RDS PostgreSQL Cluster]

Measurable Results

  • Achieved 99.99% infrastructure uptime during high-volume sale events.
  • Reduced compute infrastructure overhead costs by 35% through spot instance scaling.
  • Implemented automated rollbacks and zero-downtime rolling updates.
cloud
TerraformAzureAWSTerragruntGitHub Actions

Multi-Region Infrastructure as Code

Problem Statement:

Development, staging, and production environments were provisioned manually, leading to configuration drift and deployment failures.

Proposed Solution:

Wrote modular, reusable Terraform scripts configuration mapping out security policies, VPC subnets, and compute templates across multiple cloud environments.

View GitHub RepoLive Architecture

Deployment Topology

[GitHub Actions CI/CD] -> [Terraform Plan] -> [Approval Gate] -> [Terraform Apply] | +------------------------------+ | v [State Lock: DynamoDB Table] [State Config: Amazon S3 Encrypted Bucket] | +-------------------------+-------------------------+ | (Multi-Provider deployments) | v v [AWS Landing Zones] [Azure Resource Groups] (VPC, IAM Roles, CloudTrail) (VNet, KeyVaults, Entra ID)

Measurable Results

  • Reduced environment creation time from 4 days to 12 minutes.
  • Eliminated environment configuration drift completely.
  • Secured infrastructure states by storing secrets in HashiCorp Vault.
ai
PythonOpenAI APILangChainGrafana LokiFastAPI

AI-Powered Infrastructure Log Analyzer

Problem Statement:

Operations engineers spent hours scanning thousands of lines of logs during incident responses to isolate root causes.

Proposed Solution:

Created a background tool using LangChain and local LLMs that fetches raw Prometheus metrics and Loki logs, runs semantic searches, and generates structured root-cause reports.

View GitHub RepoLive Architecture

Deployment Topology

[K8s App Pods Log Output] -> [Vector Logging Daemon] -> [Grafana Loki Aggregator] | +----------------------------------------------------------------+ | v (Poll Loki Engine) [FastAPI Log Collector] -> [LangChain Parser] -> [Chroma Vector DB (Embeddings)] | v (semantic query matching) [Local LLM / OpenAI] | v (post report summary) [Slack Webhook Alerts]

Measurable Results

  • Reduced Mean Time to Resolution (MTTR) by 78%.
  • Automatically flagged anomalous logs that missed traditional threshold alerts.
  • Generated readable post-incident reports instantly.
cloud
AWS Control TowerAWS IAMCloudFormationTerraform

Secure AWS Infrastructure Pipeline

Problem Statement:

Internal development teams lacked standard cloud accounts, causing security exposures and lack of compliance with IAM rules.

Proposed Solution:

Set up an automated AWS Landing Zone with AWS Organizations, Control Tower, and automated security guardrails.

View GitHub RepoLive Architecture

Deployment Topology

[AWS Organizations Root] -> [Control Tower Account Factory] | +------------------------+------------------------+ | | v v [Security Tooling Account] [Log Archival Account] (GuardDuty, Security Hub) (Encrypted Audit S3 Store) | | +------------------------+------------------------+ v [Developer Sandbox Accounts] (IAM Permission Boundaries)

Measurable Results

  • Enforced 100% compliance with CIS Benchmarks across 12 child AWS accounts.
  • Automated IAM role creations with strictly bounded permissions.
  • Created centralized security auditing logs stored in a read-only S3 bucket.
devops
PrometheusGrafanaVectorElasticsearchSlack integrations

Enterprise Observability Dashboard

Problem Statement:

Lack of unified logging and metrics made performance debugging a complex task across 40 microservices.

Proposed Solution:

Integrated Prometheus agents and Vector log forwarders to pipe data to Grafana, establishing dynamic system monitoring panels.

View GitHub RepoLive Architecture

Deployment Topology

[Node Exporter Metrics] [App Custom Metrics] [Kube-State Metrics] | | | +------------------------+-----------------------+ | v (Scraped periodically) [Prometheus Server TSDB] | v (Direct DataSource) [Grafana Visual Panel] | v (Evaluates alerts) [Slack / PagerDuty Routing]

Measurable Results

  • Centralized metrics visibility in a single unified dark-mode dashboard.
  • Created predictive alerts for CPU, RAM, and disk utilization.
  • Provided developers with visual flame graphs and distributed traces.
ai
LangGraphOpenAIChromaDBNext.jsTypeScript

Enterprise Generative AI Assistant

Problem Statement:

Internal workers spent significant time searching through complex company wiki documents and PDF guides.

Proposed Solution:

Built a RAG chatbot connecting confluence wikis to a vector database, exposing a fast chat web interface for queries.

View GitHub RepoLive Architecture

Deployment Topology

[Client Application] -> [FastAPI Server Webhook] -> [LangGraph Agent Routing] | +-----------------------------+ | v [ChromaDB Knowledge Vectors] [Structured Tool Integration (MCP)]

Measurable Results

  • Answered employee infrastructure policy questions in under 2 seconds with citations.
  • Integrated Model Context Protocol (MCP) to trigger simple terminal commands securely.
  • Drastically reduced junior developer onboarding support calls.