Rohi Rikman
AI, automation & securityPractical systems. Real problems.
I work on the gaps between tools and real life—building practical systems with AI, automation and security.Where I focus
The problems connect; the tools depend on what the situation needs.Making AI useful in practice
I explore ways to connect agents, tools and useful context, with clear responsibilities and human review where it matters.
Explore this workConnecting systems and automating work
I connect systems and build focused tools around repetitive work, with visible failures and a handoff someone else can understand.
Explore this workMaking security and change manageable
I work on identity, devices and migrations, treating rollout, recovery and evidence as part of the change—not an afterthought.
Explore this workLatest writing
All articles- Why I Built OpenLTM Around Memory I Can MaintainA memory can be relevant and wrong at the same time. The choices I made in OpenLTM for that problem: a local SQLite store, an explicit lifecycle for aging and correcting knowledge, and one database shared across agent hosts, with the limits of each.
- The Agent Finished. The Work May Not Be Accepted.Herald's monitor and receipt code keeps three answers apart: what the target run did, whether anyone checked the artifact, and whether the owner received it. A walk through the fields that hold them.
- An Image Cooldown Is Not a Migration PlanMy homelab's Watchtower waits 72 hours before applying a new image, databases included. What that delay controls, and what it cannot: a PostgreSQL major upgrade, or an n8n and runner pair that has to match.