How Boost Insurance Eliminates Data Breach Risk with Pulumi

The business challenge

Boost Insurance operates in one of the most heavily regulated industries in the United States. Every policy the platform issues or binds — even a test policy — becomes a matter of record that must be reported to regulatory bodies. That means production data can’t be used for testing, and reproducing production issues in a development environment is often impractical.

Developers needed access to production data to debug real-world issues, and the only option was to download database snapshots to their laptops. For a 100 GB quoting database, that meant 45-minute downloads for remote team members, hundreds of gigabytes of sensitive data stored on personal machines, and no guarantee that data stayed within the company’s security perimeter.

“Someone goes and steals a developer’s laptop — we now have a reportable breach,” said Richard Genthner, CISO at Boost Insurance. “We have to report not only to our partners, but to the state of New York, all 50 states, and the federal government. We have to notify every single person that’s ever had a policy with us. It becomes a massive problem.”

Boost has developers working remotely across multiple countries, and some partner contracts explicitly require that data never leaves the United States — a strict data residency requirement. Their existing access-control tooling handled credential management and query logging, but couldn’t solve the fundamental problem: production data was leaving the cloud and landing on laptops.

What Boost built

Boost’s platform engineering team built an internal developer platform in Go that provisions isolated, time-limited database environments entirely within the cloud. The platform uses Pulumi’s Automation API to manage infrastructure programmatically and offers developers three ways to interact with it: a web UI, a CLI, and a desktop client.

When a developer needs to debug a production issue, they log into the platform, select a database backup, and configure the environment — instance size, disk space, and a time-to-live ranging from 8 to 96 hours. The platform then provisions an isolated VPC with a Cloud SQL (PostgreSQL) instance, restores the backup using Cloud SQL’s native import API directly from a GCS bucket, and sets up Identity-Aware Proxy (IAP)-protected routes with query logging for secure access.

The entire provisioning process takes about 15 minutes. When the TTL expires, the environment tears itself down automatically.

The results

The platform addresses some of Boost’s most critical risks while delivering measurable productivity gains:

• Data breach risk eliminated. Production data no longer leaves the cloud. Every isolated environment lives within Boost’s controlled GCP infrastructure, and data residency requirements are satisfied by default. The scenario that kept leadership up at night — a stolen laptop triggering a multi-state regulatory notification — is no longer possible.

• 30 minutes saved per developer per debugging session. Restoring a database in the cloud takes about 15 minutes, compared to 45 or more minutes for a remote developer to download and restore locally. Across Boost’s development team, that adds up to significant recovered engineering time.

• Self-service replaces an operational bottleneck. Before the platform, developers had to wait for the platform engineering team to manually provision database environments. Now developers create, extend, and tear down isolated environments on their own, freeing up the platform team to focus on higher-value work.

• Audit compliance built in. Every database connection is routed through IAP-protected tunnels with full query logging, maintaining the audit trail that insurance regulators require — the same capability that the previous tooling provided, now integrated directly into the platform.

How Pulumi makes it work

Boost selected Pulumi because its feature set aligned with the specific requirements of a self-service, compliance-focused platform.

The Automation API is the foundation. By embedding Pulumi directly in their Go application, the team could deliver a polished self-service experience — a web UI, a CLI, and a desktop client — all backed by the same infrastructure-as-code logic, without asking developers to learn Pulumi themselves.

Pulumi Deployments and its TTL stacks feature keep costs predictable and close a security gap by making every environment temporary by design. Each isolated database is provisioned with a configurable lifetime, and when it expires Pulumi Deployments tears down the stack and all of its resources automatically — so no forgotten environment lingers to accumulate cost or hold sensitive data.

Pulumi ESC (Environments, Secrets, and Configuration) shrinks Boost’s attack surface by eliminating long-lived service account keys. It issues short-lived, per-project credentials across Boost’s multi-project GCP environment using OIDC federation, and Pulumi Deployments consumes them automatically when provisioning and destroying environments — keeping the credential lifecycle fully automated.

Pulumi Cloud gives Boost’s compliance team the audit records they need without building custom logging infrastructure — exportable logs of every change across the platform. It also provides centralized visibility into every active environment, with stacks, resources, and update history in one place.

Conclusion

By building on Pulumi’s infrastructure platform, Boost turned a serious compliance liability into a self-service capability its developers actually enjoy using. Production data stays inside Boost’s controlled GCP infrastructure, isolated environments spin up in about 15 minutes and tear themselves down on schedule, and the audit trail regulators require is captured automatically — all without needing to file a request to the platform team.

For any organization that has to reconcile fast developer access with strict data-handling requirements, Boost’s platform shows the two aren’t at odds. With infrastructure as code as the foundation, self-service and compliance reinforce each other rather than compete — freeing the platform team to keep raising the bar instead of fielding one-off provisioning requests.