Restricted-hours activity creates an observation with time, site, people, vehicles, movement, and evidence.
Case study · Construction intelligence
JobLog
A motion alert tells you that something happened. JobLog is being built to manage what happens next: determine whether a person belongs there, preserve the evidence, notify the right people, track the response, and keep the case open until the outcome is documented.
- Status
- Active product development
- V1 focus
- Security incidents and site evidence
- Operating rule
- An alert is not a resolution
- Long horizon
- Construction record through building operations
The product rule
A JobLog security incident is a managed case, not a motion alert.
Detection starts the record. Closure comes later, after identity, activity, property impact, response, police involvement when applicable, and financial outcome have been documented. A person leaving the camera's field of view does not make the incident disappear.
V1 operating model
Track. Record. Notify. Resolve.
The first product promise is deliberately narrower than the long-term vision: turn an actionable site event into a structured incident, give a responder enough evidence to act, and preserve the full path to resolution.
Check-in and identity context determine whether a legitimate visitor should become a security incident.
The configured response group gets the facts and evidence needed for an immediate decision.
The case retains movement, footage, acknowledgments, notes, response actions, and escalation while it remains active.
Closure records the final classification, property impact, police or insurance involvement, and required follow-up.
The 2:00 a.m. workflow
Two people enter a material-storage area after hours.
This is the V1 security workflow expressed as an operating story rather than a specification.
Someone is there.
JobLog detects one or more people during restricted hours and first asks whether they completed the site's required check-in process.
Do they belong?
A designated check-in point can compare the visitor with the project's authorized-person list. A legitimate match records who entered, when, where, and supporting evidence without opening an unauthorized-person incident.
If not, open the case.
An unidentified person—or someone who skipped check-in—creates an active incident containing the best images, video context, people count, movement, vehicle information when visible, and areas or property approached.
Send enough information to decide.
Routing can vary by project, time, severity, and incident type. The first alert is designed to carry the decision context rather than force a responder to hunt through the application.
Keep watching the case, not just the camera.
While active, JobLog is designed to retain relevant footage, last-known location, significant activity updates, notification acknowledgments, notes, decisions, and response actions.
Document how it ended.
The outcome may be legitimate access, police involvement, theft or damage, an insurance claim, or an unresolved administrative closure. The original alert remains part of the record.
Resolution paths
Different endings. One auditable record.
The person is later confirmed legitimate. The record preserves who verified them, why they were onsite, why check-in failed, and whether access procedures changed.
Law-enforcement contact, arrival, report or case number, arrest status, and exported evidence become part of the incident history when applicable.
Stolen or damaged property, estimated value, identifying information, before/incident evidence, recovery status, and responsible contacts are retained.
When property remains unrecovered, the defined workflow can assemble the incident summary, timeline, clips, stills, inventory, estimated loss, police information, and response history into a claim package.
A case can close without an identification or confirmed loss, but it remains classified unresolved with the evidence and identifying observations retained for future correlation.
V1 architecture model
Evidence has a path. So does responsibility.
This is an architecture view, not a screenshot of a finished application. It shows the contracts and operating flow the V1 product is being built around.
Observation sources
Fixed camerasVideo + best-frame evidence
Check-in / accessIdentity + authorization context
Operator / sensorManual and connected-system observations
V1 simulatorSynthetic detections remain distinguishable
Stable seam
ObservationEvent Site + deviceTimestampPerson / vehicle dataEvidence referenceConfidence + movement
Security core
ObservationProvenance + source state
AlertPolicy + routing decision
IncidentPriority · state · owner · next action
Incident record
EvidenceImages · clips · documents · provenance
Timeline + auditAutomation, human, and external actions
NotificationDelivery · acknowledgment · escalation
ResolutionOutcome · closure · follow-up
Operating surfaces
Incident detailEvidence first, then administration
Mobile responderAcknowledge · decide · act
Reports / packagesSecurity · police · insurance · audit
V1 boundarySecurity incident management, evidence, notification, investigation, resolution, reporting, and audit. Research horizonAutonomous drones, richer world models, construction-state reasoning, and long-term building intelligence.
Evidence boundary
What is specified, what is being built, and what remains research.
JobLog has a deliberately large horizon. The credibility of that horizon depends on keeping product state explicit.
Established foundation
Defined product behavior
- V1 security-incident workflow specified from detection through closure
- Check-in, authorization, incident creation, evidence, notification, investigation, and resolution behavior defined
- Response actions and closure requirements defined as timestamped, attributable history
- Resolution paths cover legitimate access, police response, theft or damage, insurance, and unresolved closure
In development
Production platform
- Backend implementation and Azure deployment work
- Camera/video ingestion beginning with standard interfaces such as RTSP and ONVIF
- Incident workflows, evidence handling, dashboards, and operational reporting
- Identity and authorization workflows for legitimate site personnel
Research direction
Autonomous site understanding
- Local and cloud world models of the jobsite and building
- Autonomous drone capture and mission execution
- Construction-state recognition, progress deltas, and blocking-work analysis
- Long-term building intelligence extending beyond project completion
The long horizon
The construction record should outlive construction.
A building may exist for generations while much of the knowledge created during construction disappears when the project team leaves. JobLog's longer-term direction is to preserve what is there, where it is, what changed, what failed, and the evidence behind that history.
Cameras, drones, access events, people, and connected systems produce observations.
Relevant material is retained with time, location, source, and incident context.
The platform maintains an evolving representation of people, assets, work, spaces, and known site state.
Rules, workflows, operators, and eventually AI determine what deserves action.
Notify a person, assign work, preserve evidence, dispatch a mission, or update the record.
Selected work
The camera is the beginning of the story.
JobLog's V1 thesis is operational: evidence is only valuable when it becomes a managed decision and a documented outcome. The longer-term construction-intelligence vision grows from that discipline rather than replacing it.
Return to selected work →