By Jonathan Knight, Head of DevOps & Technical Manager at Utah Tech Labs. Originally published on utahtechlabs.com.
TL;DR: a production-ready AWS baseline for a startup comes down to seven controls: company-owned account, vaulted root with hardware MFA, one IAM identity per engineer with least privilege, MFA enforced by policy, AWS Budgets with tiered alerts, environment tagging, and scheduled cleanup of idle resources. This post walks through each one with the commands and configs we use.
Why rushed AWS setups become technical debt
Most early-stage AWS accounts are built through the console in a weekend: one region, one shared admin login, a couple of EC2 instances and an RDS database. It ships. Six to twelve months later the same account produces runaway bills, fragile deployments and security gaps that are painful to untangle with customers on the platform.
At Utah Tech Labs, our DevOps team architects and manages AWS for startups and growing companies, whether our developers build the product or the client's do. The goal is a foundation that is locked down from day one, cost-guarded, and designed to scale cleanly.
1. Account ownership and collaborator access
Greenfield: the company creates its own AWS account with a company email and company billing. That keeps 100% legal and billing ownership with the business. The root user is locked with hardware MFA, and outside engineers (including ours) are provisioned as collaborators through IAM roles. From there we build the VPCs, databases and CI/CD pipelines so either team can ship immediately.
Existing account: no rebuild. The client grants scoped collaborator access through an IAM role, and we audit for security holes, untagged resources and single points of failure, then harden in place without disrupting live services.
A useful first audit command lists every IAM user and whether they have MFA devices:
aws iam generate-credential-report aws iam get-credential-report --query Content --output text | base64 --decode \ | cut -d, -f1,4,8 | column -t -s, # columns: user, password_enabled, mfa_active Any human user with password_enabled=true and mfa_active=false is your first fix.
2. Zero-trust access baseline
Three rules, no exceptions.
Rule 1: vault the root user. Root is never used for deployments or maintenance. Enable a hardware MFA device on it, remove any root access keys, and store the credentials offline. Check for root access keys with:
aws iam get-account-summary --query "SummaryMap.AccountAccessKeysPresent" # should return 0 Rule 2: one identity per engineer, least privilege. No shared logins. Every engineer on either team gets an individual IAM identity scoped to what they actually need. Offboarding becomes "delete one user", and CloudTrail shows who did what.
Rule 3: enforce MFA by policy. A guideline is not enough; deny access when MFA isn't present. This pattern, attached to a group that every human user belongs to, blocks everything except the actions a user needs to set up their own MFA device:
{ "Version": "2012-10-17", "Statement": [ { "Sid": "DenyAllExceptMFASetupWithoutMFA", "Effect": "Deny", "NotAction": [ "iam:CreateVirtualMFADevice", "iam:EnableMFADevice", "iam:ListMFADevices", "iam:ResyncMFADevice", "iam:GetUser", "iam:ChangePassword", "sts:GetSessionToken" ], "Resource": "*", "Condition": { "BoolIfExists": { "aws:MultiFactorAuthPresent": "false" } } } ] } Enforcing MFA across every user account wipes out over 90% of automated credential attacks before they become a threat. Test policies like this on a non-admin test user first so nobody gets locked out.
3. Cost governance: budgets, tagging and cleanup
Budgets with tiered alerts. We create AWS Budgets at thresholds that match the company's stage, for example $100, $500 or $1,000 a month, and alert on forecasted spend so the warning arrives before the overrun. Alerts go to email and Slack (via SNS).
budget.json:
{ "BudgetName": "monthly-500", "BudgetLimit": { "Amount": "500", "Unit": "USD" }, "TimeUnit": "MONTHLY", "BudgetType": "COST" } notifications.json:
[ { "Notification": { "NotificationType": "FORECASTED", "ComparisonOperator": "GREATER_THAN", "Threshold": 80, "ThresholdType": "PERCENTAGE" }, "Subscribers": [ { "SubscriptionType": "EMAIL", "Address": "alerts@yourcompany.com" } ] } ] aws budgets create-budget \ --account-id 111122223333 \ --budget file://budget.json \ --notifications-with-subscribers file://notifications.json Environment tagging. Every resource carries an Environment tag (dev or prod) so the bill splits cleanly. Activate the tag as a cost allocation tag in the Billing console, then you can find untagged or mis-tagged resources with the Resource Groups Tagging API:
aws resourcegroupstaggingapi get-resources \ --tag-filters Key=Environment,Values=prod \ --query "ResourceTagMappingList[].ResourceARN" Scheduled cleanup. Unattached EBS volumes and unassociated Elastic IPs keep billing after anyone needs them. These two queries find them:
# EBS volumes not attached to any instance aws ec2 describe-volumes --filters Name=status,Values=available \ --query "Volumes[].{ID:VolumeId,SizeGiB:Size,Created:CreateTime}" --output table # Elastic IPs not associated with anything aws ec2 describe-addresses \ --query "Addresses[?AssociationId==null].PublicIp" --output text We run checks like these on a schedule and review before deleting, rather than deleting automatically.
4. Third-party tooling that keeps the stack simple
- Cloudflare in front of the application for instant SSL and DDoS protection.
- Datadog or Sentry for monitoring and error tracking, so developers see errors in real time instead of grepping raw CloudWatch logs.
Fewer moving parts inside AWS means faster debugging and fewer late-night incidents.
Common mistakes we see in AWS audits
When we audit an existing account, the same patterns come up again and again. None of them are exotic, and all of them are cheaper to prevent than to fix.
The account belongs to a former contractor. The root email is a personal Gmail or an agency address, and the billing card is theirs. Recovering ownership later is slow and awkward, especially during a funding round or a security questionnaire.
One admin user shared by the whole team. It's convenient on day one and a liability on day ninety. There is no way to tell who changed what, and removing a departing engineer means rotating credentials for everyone.
MFA "recommended" but not enforced. Some users enable it and some don't. The accounts without it are exactly the ones that automated credential attacks find.
No budget, or a budget nobody reads. Alerts go to an inbox nobody monitors, or the only alert fires on actual spend at 100%, after the money is gone. Forecasted alerts to a shared channel fix this.
No tags, so no cost attribution. When the bill jumps, nobody can say whether it was production traffic, a forgotten load test, or a dev database left running over a holiday.
Orphaned storage and IPs. Terminated instances leave behind EBS volumes and Elastic IPs that quietly bill every month.
Each of these maps directly to one of the controls above, which is why we apply them as a set rather than one at a time.
The checklist
| Control | Done when |
|---|---|
| Company-owned account | Account email and billing belong to the company |
| Root vaulted | Hardware MFA on root, zero root access keys |
| Individual IAM identities | No shared logins; least-privilege access |
| MFA enforced | Deny-without-MFA policy on all human users |
| Budgets | Tiered forecasted alerts to email and Slack |
| Tagging | Environment tag on every resource, active for cost allocation |
| Cleanup | Scheduled review of unattached volumes and idle IPs |
FAQ
What's the minimum AWS security baseline for a startup?
A company-owned account, a vaulted root user with hardware MFA, individual least-privilege IAM identities, MFA enforced by policy, and audit logging so every action is attributable.
How do I enforce MFA for all IAM users?
Attach a policy that denies all actions except MFA self-setup when aws:MultiFactorAuthPresent is false, to a group that every human user belongs to.
How do I get alerted before an AWS bill overruns?
Create an AWS Budget with a FORECASTED notification at a percentage threshold, with email and SNS subscribers.
How do I find idle AWS resources that cost money?
Query EBS volumes with status available and Elastic IPs with no AssociationId, then review and remove them on a schedule.
Can an existing AWS account be hardened without downtime?
Yes. With scoped IAM role access, you can audit and fix access, tagging and cost controls in place without disrupting live services.
Want a second pair of eyes on your account?
Utah Tech Labs offers a free AWS review covering ownership, access, cost controls and resilience. We'll tell you plainly what to fix first.
More on our DevOps & Cloud Engineering services, or email grow@utahtechlabs.com.
Jonathan Knight is Head of DevOps & Technical Manager at Utah Tech Labs.

