AWS Cloud Computing Training in Pune with Placement

Cloud computing concepts (IaaS / PaaS / SaaS, on-prem vs cloud trade-offs, shared responsibility model), the AWS global infrastructure (Regions, AZs, Edge locations, Wavelength zones), account creation, billing alarms, the AWS Free Tier, and IAM basics. By the end of week 1 every student has a personal AWS sandbox account, MFA enabled, a budget alarm at ₹1,000, and an IAM user with programmatic access. We deliberately spend a full session on cost — every horror story you have heard about a runaway AWS bill starts with what we cover this week.

Identity is the foundation of cloud security. Learn IAM users, groups, roles, and policies (managed, customer-managed, inline) — and the design principles that separate hobbyist accounts from production-grade ones. Cover IAM Identity Center (the successor to AWS SSO) for workforce identity, OIDC federation for CI/CD pipelines, and AWS Organizations + SCPs for multi-account governance. We finish the module by hardening the sandbox account using the AWS Security Hub baseline.

Cloud networking is where most certification candidates lose marks and most engineers lose production weekends. We build a VPC from scratch — public and private subnets across two AZs, a NAT Gateway, an Internet Gateway, route tables, security groups, and Network ACLs — and prove the design with a working three-tier application (ALB → EC2 → RDS).

CIDR planning gets its own session, because a VPC range chosen badly in week one is the thing you cannot change later without rebuilding. We size subnets against realistic growth, reserve space for future AZs, and check the plan against the on-prem ranges an enterprise would need to peer with.

The second half of AWS networking — connecting a VPC to other VPCs, to on-premises, and to AWS services without traversing the public internet. VPC Peering and its non-transitive limitation, Transit Gateway as the hub-and-spoke answer, Gateway and Interface endpoints, PrivateLink for exposing your own services, and Direct Connect versus Site-to-Site VPN.

Route 53 closes the module: hosted zones, record types, and the routing policies (weighted, latency, failover, geolocation) that appear in both the SAA-C03 exam and every real disaster-recovery design.

The compute service everything else is measured against. EC2 instance families and what each one is actually for, the four pricing models (on-demand, spot, savings plans, reserved) and how to combine them, storage-optimised versus compute-optimised selection, AMI lifecycle and EC2 Image Builder, and user data for bootstrapping.

Then resilience: Auto Scaling groups with launch templates, target tracking and step scaling policies, health checks, and an Application Load Balancer in front. The lab ends with an application that survives an instance being terminated by hand.

Serverless as an architecture, not a trick. Lambda in depth — runtimes (Java 21, Python 3.13, Node 22), layers, memory-to-CPU coupling, timeouts, concurrency limits, and the cold-start behaviour that decides whether serverless suits a workload at all. Then the services Lambda is glued to: API Gateway REST and HTTP APIs, Step Functions for orchestration, EventBridge for routing, and SQS / SNS for decoupling.

The lab builds a small event-driven service end to end, so the trade-off against the EC2 design from the previous module is something you have measured rather than read about.

Where most Pune cloud roles actually spend their time. Push images to Amazon ECR with scanning enabled, then run them two ways: Amazon ECS with Fargate (task definitions, services, service discovery, autoscaling) and Amazon EKS (managed node groups, Fargate profiles, the AWS Load Balancer Controller, IRSA for pod-level IAM).

We are explicit about when each is the right answer — ECS for a team that wants AWS to hold the abstraction, EKS for a team that needs the Kubernetes ecosystem or portability. You build a Fargate-only EKS cluster you can run for under ₹100 a day.

S3 in the depth the exam and the job both require: storage classes and the retrieval cost that makes Glacier cheap only if you never read it, lifecycle rules, versioning and MFA delete, replication (same-region and cross-region), encryption modes, signed URLs, static website hosting, and S3 Object Lambda.

Block and file storage follow — EBS volume types and the gp3 baseline-versus-provisioned distinction, snapshots and fast snapshot restore, EFS for shared POSIX filesystems, and an overview of FSx. The module closes on choosing correctly between the three under a stated cost and latency budget.

Managed data services and the design decisions that outlast them. RDS engines, Multi-AZ for availability versus read replicas for scale (a distinction interviewers reliably probe), parameter groups, backups and point-in-time recovery. Aurora goes deeper — the shared storage layer, Serverless v2 scaling, and Aurora Global for cross-region.

DynamoDB is taught as a modelling discipline rather than a key-value store: partition key design, hot partitions, single-table design, GSIs and LSIs, on-demand versus provisioned capacity, and Streams. ElastiCache and OpenSearch close the module.

Modern cloud is code, not clicks. Terraform 1.7+ first — providers, resources, variables and outputs, remote state in S3 with DynamoDB locking, modules, workspaces, `plan` as a review artefact, and the import workflow for adopting resources someone created by hand. Then AWS CDK in TypeScript — constructs, stacks, and how synthesis produces CloudFormation underneath.

We teach both because Pune teams use both and you will be asked to read each. CloudFormation itself is covered at the level needed to debug what CDK emits: stacks, change sets, and drift detection.

Wiring the code you write to the infrastructure you defined. GitHub Actions deploying via OIDC federation — no static access keys anywhere, which is now the expected answer in interviews — plus AWS-native CodePipeline, CodeBuild and CodeDeploy for shops standardised on AWS tooling.

Deployment strategy is the second half: rolling, blue/green with ALB target-group swaps, and canary releases with CodeDeploy traffic shifting. Every pattern is paired with its rollback path, because a deployment strategy without a rollback plan is just optimism.

Observability the way SREs actually configure it, not the way marketing pages describe it. CloudWatch metrics and custom metrics, structured JSON logging queried with Logs Insights, alarms with sensible thresholds feeding SNS, composite alarms to suppress noise, and dashboards that a person on call can read at 3am.

Distributed tracing follows with X-Ray and the AWS Distro for OpenTelemetry, plus a small Prometheus and Grafana stack on EKS for the cloud-native pattern you will meet in container-first teams.

The SAA-C03 exam weights security heavily, and so do real Pune hiring panels. Cover KMS in depth (CMKs, grants, envelope encryption, key rotation), AWS Secrets Manager and Parameter Store, ACM for TLS certificates, AWS WAF and AWS Shield (Standard and Advanced), Macie for data classification, and the Well-Architected Tool's six pillars.

We finish by reviewing each student's capstone architecture against the Well-Architected lens — a written review, in the format AWS partners actually produce.

The skill that gets a cloud engineer promoted and the one most syllabi omit. Reading a bill properly with Cost Explorer and the Cost and Usage Report, allocating spend with tagging strategies and cost categories, and setting Budgets with actions that actually stop runaway resources rather than merely emailing about them.

Then the levers: right-sizing from CloudWatch data, Savings Plans versus Reserved Instances, spot for interruptible work, S3 lifecycle and storage-class analysis, and the standard waste list — idle load balancers, unattached EBS volumes, orphaned snapshots, over-provisioned RDS.

Pune cloud architects in 2026 are expected to design AI features, not just classic three-tier apps. Cover Amazon Bedrock (Claude, Llama, Titan, Mistral, Stable Diffusion models on a single API), Bedrock Guardrails for content filtering, Bedrock Knowledge Bases backed by Amazon OpenSearch Serverless or Aurora pgvector, and Bedrock Agents for tool-using assistants. Build a small retrieval-augmented generation (RAG) service — S3 PDFs → Knowledge Base → Bedrock model → API Gateway endpoint — that you can demo end-to-end in interviews.

Two weeks of full-time capstone work plus structured Solutions Architect Associate (SAA-C03) preparation. Pick one of three capstone architectures (see Capstone Projects). For exam prep we run question-bank drills, scenario walkthroughs, and two full-length mock exams under timed conditions — students typically score 75–85% on the second mock if they have done the lab work seriously. Mock interviews target Pune cloud hiring panels (Persistent, BMC, Bajaj Finserv, Synechron). Resume / LinkedIn / GitHub polish included.