Saad Mahboob
All case studies

Fleet management

The product I own now: deploying, running and fixing a robot fleet

Every robot that ships needs deploying, testing, supporting and fixing. I own fleet management at Sereact: the tools that make that faster, and the site fixes we turn into platform features so every customer gets them.

Role
Product owner, fleet management
Company
Sereact
Scope
Deployment, operations, support and serviceability
Team
Also stood in as lead of a core robot software team
  • Weeklyrelease train, from merge to customer fleet
  • 2–5 minfrom ticket to root-cause report
  • 3customer deployments set up with the wizard, without a robotics engineer

The release train: from merge to fleet

I saw problems in how releases reached customers, so I designed a weekly release train and automated it end to end. One click sends a tested release to Customer Success, and once they give the go-ahead it ships on release day.

  1. Every merge

    Checks

    Every module is built and checked automatically.

  2. Nightly

    Simulation

    Every candidate is tested in simulation first.

  3. Nightly

    Real robots

    Then on real robots, overnight.

  4. Weekly

    Freeze

    A clean candidate and its release notes, picked automatically.

  5. Weekly

    Sign-off

    Customer Success signs off with customers.

  6. Release day

    Roll out

    Canary robots first, then the fleet.

  7. After release

    Watch

    Automatic rollback if anything degrades.

Bugs that reach customers become new simulation tests, so they can’t come back.

Who it’s for

The people who put robots on a customer’s floor and keep them there: deployment engineers, Customer Success, the operators at each station, and the partners who now install stations for us.

The problem

The fleet was growing faster than the way we ran it. Setting up a station took a robotics engineer on site. Releases reached customers by hand, one site at a time. Every support ticket started with an engineer reading logs. None of that scales to 100 stations.

Deploy faster

  • A one-screen rollout console for the production fleet, now running two product lines.
  • A deployment wizard, so stations can be set up without a robotics engineer. I started it and a colleague built it; it has been used on three customer deployments.
  • Launched the move to suppliers deploying stations end to end, together with four colleagues.

Run the fleet

  • An AI support agent, built with two colleagues, that turns a ticket into a root-cause report in 2–5 minutes.
  • A site daily report: a five-minute morning read per site on picks, KPIs and lost output, built from feedback I got from customers directly.
  • Technical lead for running a vision-language model on premises, live on every pick cycle in production.

Fix it before it breaks

  • Parts and visit tracking, live in Vienna, so every service visit is accounted for.
  • A one-tap maintenance log on the station screen, built from an operator’s feedback.
  • Picks per intervention as the one number that ties the serviceability work together.

What changed

  • Setting up a station went from a robotics engineer on site to a guided wizard. It has been used on three customer deployments.
  • Releases went from by hand, site by site, to one click: tested overnight, signed off by Customer Success, shipped weekly.
  • Support went from an engineer reading logs to a root-cause report in two to five minutes.
  • 9+ site acceptance tests passed on site, and two stations live in three days with acceptance passed in five.

What I learned

  • The fix that matters is the one that ships to every site.
  • Measure before you ask. Data moves a roadmap faster than a complaint.