About
Built by a clinician, for clinicians.

Michael Olszewski
PharmD, BCPS, BCCCP
Clinical Pharmacist Specialist
Clinical AI & DevOps Engineer
Charlotte area, NC
AWS Certified Solutions Architect – Associate (SAA-C03)In progress
The founder
Two decades in acute care, now building the systems behind it.
Michael Olszewski is a dual board-certified clinical pharmacist with two decades in critical care and infectious disease pharmacy. For eight years he led an antimicrobial stewardship program. He is also the engineer who builds LocumView: the identity layer, the desktops, the cluster they run on, and the documentation behind every change.
LocumView comes from the gap between those two jobs: the extra logins, the copied numbers, the alerts that arrive too late or too often.
The systems thinking came first. At ten he was writing BASIC on a 286 running MS-DOS, dialing into BBS services and opening up machines to see how the hardware worked. Linux has been part of his life ever since, long before it was part of his work.
Critical care made the cost of bad software concrete: a lab value copied by hand from one screen to another, an alert that fires on every patient until nobody reads it, one more login when there is no time for one. Stewardship taught the other half: clinicians change practice when they are shown evidence they can check, not when they are told to.
So every part of LocumView answers to two reviewers. The pharmacist asks whether a suggestion is safe, sourced and worth a clinician's attention. The engineer asks whether the system can be rebuilt from code, audited after the fact and run without asking a vendor's permission.
He works the way a pharmacist verifies an order: nothing goes live unchecked, and every change is written down. The architecture decisions, the changelog and the roadmap are public, including the parts that are not done yet.
His focus is agent-native clinical workspaces for regulated healthcare. The EHR becomes a governed data source over FHIR. Above it sits one unified environment: a hardened Linux desktop with AI agents built into the operating layer, centralized dashboards, and workflows tailored to each type of clinician. Everything is designed to run inside the regulated boundary, so clinicians work in one place instead of a dozen tabs.
He designs end to end, from network boundary and identity to workload placement and the clinical workflow on top, and is working toward the AWS Certified Solutions Architect – Associate (SAA-C03) certification for cloud and hybrid deployments.
Related clinical AI work
- AMS Intelligence(opens in a new tab)Antimicrobial stewardship decision support that runs entirely on local models. Powers resistance trend analysis, antibiotic utilization signal detection, guideline-grounded clinical Q&A, and automated stewardship reports, with confounding by indication explicitly modeled.FAERSNHSNWHONETATLASPubMedIDSA / ASHP guidelines
- PV AI Workbench(opens in a new tab)Pharmacovigilance signal detection and RAG-based regulatory Q&A, plus MedDRA coding, ICSR narrative drafting, and literature monitoring. All inference runs locally.Retrieval: 100% Precision@5, MRR 0.861Signal detectionRAGMedDRAICSR
Stack
- Python
- Bash
- RHEL & SELinux
- KVM / libvirt
- bootc (RHEL image mode)
- Podman
- Kubernetes (k3s)
- Terraform / OpenTofu
- Ansible
- Keycloak & Red Hat IdM
- OpenSCAP / DISA STIG
- SOPS / age
- CI (Gitea Actions)
- Ollama (local LLMs)
- LangChain & ChromaDB
- FHIR
Self-hosted infrastructure
- Lab
- Self-hosted Kubernetes (k3s) with local LLM inference on GPU
- Practices
- Infrastructure as code, versioned changes, encrypted off-site backups
I design systems that clinicians can trust and engineers can maintain.
Vision
What we are building toward.
A clinical workspace where the right information arrives before it is needed, where every automated suggestion shows its evidence, and where IT can rebuild any environment exactly, on infrastructure they control.
We believe open, enterprise Linux is the right foundation for that: auditable, portable, and free of lock-in.
Clinician in control
Software drafts and flags. People decide.
Show the evidence
Every suggestion cites its data and its sources.
Build from code
If it cannot be rebuilt, it cannot be trusted.
Open foundations
Standards-based, portable, and not tied to one vendor.