Matheus Castelliano
← All work

Product owner2026 · Network software lifecycle

From “you’re vulnerable” to “you’re patched”.

For a major telecom operator in Brazil: know which switches are at risk, decide which version they need, and upgrade them safely.

Role
Product owner
Domain
Network infrastructure
Users
Network operations teams

One flow, from diagnosis to action

Know. Decide. Act.

  1. 01

    Know

    Match known vulnerabilities (CVEs) to the real devices in the inventory.

  2. 02

    Decide

    Recommend the target software version for each device.

  3. 03

    Act

    Upgrade in scheduled batches, with checks before and after.

  • Pre-check
  • Human approval
  • Software update
  • Reboot and reconnect
  • Post-check: compare before vs after
Upgraded and verified.
01

Problem

Network teams had no easy way to see which switches ran vulnerable software, which version to move to, or how to upgrade safely. Each upgrade was done by hand and took about an hour.

02

What I defined

One flow from diagnosis to action: match CVEs to real devices, recommend the target version, and run the upgrade with automatic checks before and after. Upgrades are scheduled in batches per datacenter, with progress, logs and evidence for every switch.

03

Decision: earn trust before automating

The first release required a human approval after every pre-check. Only once teams trusted the checks did we add a fully automatic mode. Routine upgrades now run on their own; risky ones still get a human in the loop.

04

Scope call

The first release left out image upload and the full check suite. Getting a safe, approvable upgrade in front of users came first; the rest followed.

05

Result

An upgrade that took about an hour by hand now runs in minutes, validated end to end in the operator’s own network: pre-check, approval, update, reboot, reconnection and post-check.

06

Trade-off

Two modes, manual approval and fully automatic, made the product more complex to build and test. Worth it: it’s what let teams adopt it at their own pace.

07

What I’d do differently

Test against real network topologies earlier. A redundancy setup (MLAG) only showed up in the client’s environment, and the pre-check had to grow after launch.

Let’s talk.

A role, a project or just a conversation about AI products. Send me a line; I answer every message.

matheus_castelliano@outlook.com

Open to JobsFreelanceConsulting