System Administrator · Atlassian AI & Automation · Stockholm, Sweden

I turn operational knowledge into reliable IT support.

I set the rules for which Confluence pages assistants can trust: an owner, a review cadence and a path to retirement. Then I test whether Atlassian Rovo and Risotto retrieve the maintained pages.

Selected Work

AI, Automation, and Operational Knowledge

Confluence Cloud · Jira Service Management · Atlassian Rovo · Risotto

Focus: I lead AI and automation initiatives across enterprise Atlassian systems at Optimizely. My focus is structuring operational knowledge lifecycles so service workflows and conversational assistants (Atlassian Rovo, Risotto) retrieve maintained, owned guidance rather than unverified documentation.
Architecture: I established metadata schemas, page ownership requirements, review cadences, and automated quarantine for obsolete documentation. Every change goes through a ticket and a sandbox test before it reaches production.
Evaluation: I test each workflow as a request path with four checks: it finds the right page, asks when context is missing, stops at an approval gate, and hands a person the diagnostic state on escalation. A fluent answer is not evidence of a correct resolution; resolving suitable requests early is the goal.

Enterprise Systems and Identity Operations

Microsoft Entra ID · Okta · Microsoft Intune · Jamf Pro

Scope: I support enterprise identity and access workflows, macOS and Windows endpoint fleet configurations, hardware inventory, and Service Desk escalations.
Approach: My focus is operational consistency: investigate issues, establish repeatable administrative baselines, automate routine provisioning, and route edge cases to documented functional owners.

Private Infrastructure and Local Inference

Linux · Apple Silicon · Virtualization · Local models

Setup: I maintain a private engineering lab to explore virtualization, networking, service isolation, and local model inference on Apple Silicon hardware.
Approach: Local models complement rather than replace frontier cloud intelligence. I evaluate how high-throughput local inference on Apple Silicon can accelerate iterative developer workflows, measuring exact runtime throughput against workload constraints and auditing data boundaries separately.
Evidence: I publish the measurements as local-model-check, an open project with the synthetic tasks, raw results and a list of what was not tested.
github.com/ZeshanNasir/local-model-check ↗

Engineering Decisions

Knowledge lifecycle

Outdated guidance misleads enterprise search and AI assistants.

Keep Confluence as the maintained source of truth. Automated review cadences quarantine unowned pages so AI tools never index stale knowledge.

Service workflows

A plausible AI answer does not prove correct resolution.

Evaluate the full request path: retrieve guidance, clarify missing context, enforce approval boundaries, and escalate with structured diagnostic state.

Local inference

Running a model locally does not guarantee zero data egress.

Measure what runs reliably on Apple Silicon hardware, and audit network traffic separately. Make no privacy claim without an empirical egress test.

Disaster recovery

A configured backup is not proof that a restore will work.

Treat backup scope and restore-test results as separate evidence. Record what was included, what was restored, and what the recovery runbook actually proved.

About

I am a System Administrator in Stockholm. This page is the record of how I work: one authoritative source, a named owner, and evidence that the guidance still holds.

The same habits run my personal lab, where I test virtualization, networking, observability and local models on hardware I own.

I prefer systems where the source of truth, ownership, operating limits and recovery status can be checked rather than assumed. I keep employer work and personal experiments distinct, and publish only material that can be shared appropriately.

Virtual University of Pakistan, B.S. in Information Technology

Uppsala University, Information Systems: 45 of 60 ECTS completed

Contact

contact@zeshannasir.com LinkedIn ↗ GitHub ↗