Detect
Monitoring, vulnerability, availability, and policy signals identify a condition.
Local AI · Governed autonomy
AI-driven IT for a clinic with no dedicated IT department.
HELIOS-1 is a working operational model for a single provider, one nurse, and one administrator—where local AI helps run the technology environment from detection through verified resolution.
Local-firstInference and operational context remain under local control.
Evidence-drivenEvery action returns proof to the workflow record.
Policy-boundedAI capability never becomes authority by default.
OPERATIONAL SIGNAL
Condition detected
Availability · security · vulnerability · policy · capacity
01ObserveCollect evidence
02ValidateEstablish scope
03AuthorizeApply policy
04ExecuteBounded action
05VerifyProve outcome
Evidence returnedTicket updated · outcome recorded · closure evaluated
AUDITABLE
SIMULATED CLINIC MODEL
01Provider
01Nurse
01Administrator
00Dedicated IT staff
THE OPERATING IDEA
It does not just recommend.
It does not just recommend.
It operates—inside policy.
HELIOS-1 joins identity, infrastructure, security evidence, workflow governance, and local AI into one controlled operating model. It can recognize conditions, assemble context, create and route work, perform approved actions, verify the result, and preserve the evidence.
The objective is not unrestricted automation. It is dependable autonomy: routine work handled consistently, uncertain work blocked safely, and consequential work escalated with context.
CLOSED-LOOP OPERATIONS
From signal to verified resolution.
Every stage has a distinct responsibility. Detection never becomes permission, and execution never becomes proof until the environment verifies the outcome.
Open
A structured ticket becomes the workflow record and audit surface.
Validate
The request, target, risk, boundaries, and success criteria are evaluated.
Assign
Approved work moves to the correct AI layer with a bounded contract.
Resolve
Only the authorized action is performed against the specified target.
Verify
Post-checks confirm whether the requested outcome was actually achieved.
Document
Evidence is posted, the ticket is closed or escalated, and history is retained.
AI ≠ AUTHORITY
Models can interpret, recommend, and generate actions. Policy determines whether an action is allowed. HELIOS-1 keeps those responsibilities separate by design.
AUTONOMOUS CAPABILITIES
Built to keep the clinic operating.
HELIOS-1 applies the same governed loop across availability, cybersecurity, maintenance, recovery, and operational hygiene.
SELF-HEALING
Service recovery
Detects a failed or degraded service, gathers evidence, opens governed work, performs an approved recovery action, and verifies that the service returned.
DetectRecoverProve
THREAT-LED
CISA KEV response
Uses daily Known Exploited Vulnerabilities intelligence to identify relevant exposure, prioritize remediation, route patching, and validate resolution.
Daily intelligenceVerified remediation
MAINTENANCE
Managed patching
Turns patch requirements into bounded change workflows with pre-checks, controlled deployment, health validation, evidence, and escalation.
Risk-awareEvidence-backed
CONTINUITY
Backup stewardship
Coordinates backup operations, retention checks, integrity evidence, failure handling, and recovery-readiness workflows across protected services.
RetentionRecovery readiness
ASSESSMENT
Vulnerability lifecycle
Finds exposed conditions, adds asset and risk context, creates tracked work, controls remediation, and uses rescanning to confirm closure.
DiscoverRemediateRescan
TRUST
Identity and policy
Maintains a centralized trust foundation for users, services, certificates, administrative access, and policy-governed change.
Least privilegeAccountability
THE AI AUTHORITY CHAIN
Different agents. Deliberately different authority.
No single model is allowed to observe a condition, authorize itself, expand its scope, and declare success. HELIOS-1 separates the work.
Operational signals
Monitoring, scanners, asset records, risk context, and availability checks provide evidence—not instructions.
P
GOVERNANCE LAYER
Pilot Agent
Validates the request and defines target, risk, allowed actions, forbidden actions, success criteria, and stop conditions.
T
CONTROL LAYER
Ticket Agent
Finds eligible work, selects the approved AI backend, prevents duplicate runs, and dispatches one bounded job at a time.
H
EXECUTION LAYER
Helios Agent
Performs only the authorized work, collects post-change evidence, updates the record, and closes or escalates the ticket.
SYSTEM ARCHITECTURE
The machine underneath the intelligence.
Public-safe infrastructure roles show how HELIOS-1 distributes trust, compute, evidence, applications, and execution without exposing private implementation details.
PROTECTED CONTROL PLANE
AI Orchestration
Context · policy · dispatch · execution · evidence
01Local-firstOperational reasoning stays under controlled infrastructure boundaries.
02SegmentedTrust, applications, security workloads, and execution remain intentionally separated.
03ReproducibleInfrastructure and workflows are documented, testable, and designed to be rebuilt.
04ObservableActions and outcomes produce evidence that can be reviewed later.
PLATFORM STACK
Real systems. Real integrations. Local control.
HELIOS-1 is assembled from proven infrastructure and security technologies, connected by a governed AI operations layer.
Local AI & orchestration
A dedicated GMKtec EVO-X2 local-inference system with AMD Ryzen AI and 128 GB of unified memory, running Qwen3.8-Flash-Next through llama.cpp with Vulkan acceleration alongside policy-aware agent orchestration.
Identity & trust
Central authentication, administrative policy, certificate services, secure directory integration, and managed endpoints.
Security & availability
Endpoint telemetry, infrastructure logs, availability detection, vulnerability assessment, and controlled validation.
Workflow & evidence
Ticket governance, asset lifecycle, risk treatment, internal notifications, application data, and auditable history.
Clinic simulation
A working healthcare-oriented application surface and managed user endpoints for provider, nursing, and administrative workflows.
Infrastructure foundation
Local virtualization, Linux services, containerized applications, reverse proxying, network segmentation, and controlled egress.
OPERATIONAL, NOT THEORETICAL
The core autonomous workflow has been proven end to end.
HELIOS-1 has demonstrated unattended ticket detection, AI-led request validation, bounded authorization, controlled execution, evidence posting, and ticket closure against real infrastructure.
- ✓ Local inference serving an operational agent workflow
- ✓ Ticket-driven separation of validation and execution
- ✓ Cross-platform Linux and Windows administration
- ✓ Database locking and serial agent dispatch
- ✓ Evidence-backed completion and ticket closure
00:00SIGNAL Condition enters governed workflow
00:01TICKET Execution record established
00:02PILOT Scope and authority validated
00:03DISPATCH Bounded run acquired
00:04HELIOS Approved action executed
00:05VERIFY Success criteria confirmed
00:06CLOSED Evidence retained
OUTCOME: CONTROLLED / VERIFIED / AUDITABLE
THE PROJECT
A working model for the future of small-clinic IT.
HELIOS-1 asks a practical question: can a carefully governed local AI platform provide dependable IT operations for a very small clinic that has no dedicated technical staff?
It is an independent research and engineering platform—not a production clinical environment, not a repository for real patient data, and not a claim of HIPAA compliance.