Project Rescue & Optimization
For software that is late, slow or failing in production, and a team that no longer knows why.
A rescue starts with reading: the code, the logs, the deploy history. Before changing anything we want to be able to say what is actually wrong, in terms you can check for yourself.
Then we fix the things that are hurting now, and give you an honest recommendation for the rest, whether that is repair, a partial rebuild or a fresh start. No judgment about how it got this way.
What we take on
Production failures
Hangs, crashes and data that goes missing. Root cause first, then a fix that is tested against the failure.
Performance
Slow pages, slow queries and hosting bills that grow faster than usage, measured before and after.
Code audits
An outside read of a codebase before you buy it, inherit it or build on top of it.
Migrations and upgrades
Off an end-of-life framework, database or host without stopping the business while it happens.
How an engagement runs
Scope
A conversation, then a short written plan: what gets done first, what it costs, and what is deliberately left out.
Build
Working results early and often. You see progress in something running, not in status reports.
Ship
Deployment, monitoring and backups are part of the job, not an afterthought.
Hand over
Your code, your accounts, and documentation written for whoever maintains it next.
What we use
Work that backs this up
Tracing a hang to its cause in a large codebase we had never seen before.
A fork-safety fix for tt-metal
Contribution to an AI hardware framework
Tenstorrent's tt-metal hung when a forked process opened a device. We traced it to state inherited across fork() and wrote the fix.