EPHOR Ai lets an airline and its civil aviation authority run training oversight from one shared record: publish the training programme, arrange inspector observations, get course material approved, and keep instructors on the approved version. Each approval is pinned to the exact files reviewed and issued as a numbered certificate of approval that both organizations receive, and every decision is made by an authorized person and kept in an append-only history.
Email me the brochure, EPHOR Ai
EPHOR Ai is FlightAtom's training oversight and approval software, shared by an airline and its civil aviation authority. The operator publishes its training programme and submits course material; authority inspectors request observations, record attendance and outcomes, and approve or return what is submitted. Each approval is pinned to the exact file versions reviewed and issued as a numbered certificate of approval with a SHA-256 fingerprint for every file.
An operator changes a training session after observation has been confirmed. The reviewer needs to see what changed and reconfirm attendance.
Update the training session. A material change marks the observation as needing reconfirmation so the inspector can review it again.
In the separate approval workflow, submit a document package. The reviewer can request changes or record a decision against its exact revision.
Inspect the approval snapshot and history to establish what was reviewed, by whom, and at which revision.
An updated observation record and a separate approval package, each retaining its changes and decisions.
The authorized reviewer makes the decision against the package revision and records any conditions or required changes.
Marking a course folder complete freezes it as a numbered version. When a person at the authority approves it, EPHOR Ai issues a certificate for that exact file set, and instructors see the course only while its current version is approved.
These EPHOR Ai screens were captured in an isolated demo with neutral organization labels and synthetic records. Open a screenshot to inspect the workflow.
The month calendar an inspector sees after the airline publishes its training programme. Each session type has its own colour, and sessions without a confirmed time show as time TBC. Synthetic demonstration data.
The airline's queue of observation requests from the authority. It holds a new in-person request to acknowledge, confirm or decline, one request being arranged, and a confirmed remote observation with its arrangements and a joining link that is kept out of email. Synthetic demonstration data.
An inspector's observations that need attention. The airline changed two sessions after confirming them, so both now wait for re-confirmation. The airline could not host a third, and it gives a reason and suggests an alternative. Synthetic demonstration data.
An authority reviewer's view of a course folder whose second version is frozen and awaiting a decision. Version 2 is compared with version 1 file by file, with Approve and Return to pending buttons. Synthetic demonstration data.
Page 1 of the numbered Certificate of Approval issued when the authority approved version 2 of a course. It shows who requested, submitted and approved the course, what the approval covers, and each approved file with its SHA-256 fingerprint. Synthetic demonstration data.
A flight instructor's library, limited to Training Material and Account. It lists only the course folders whose current version the authority has approved. Synthetic demonstration data.
The operator's programme, the inspectors' observations, the course material and every decision sit on one record that both organizations work from.
Each approval issues a numbered certificate of approval as a PDF, listing every approved file with its SHA-256 fingerprint, the approver's note and the decision history. The certificate is stored unchanged, and the same PDF is emailed to both organizations.
Instructors see only the version of each course that the authority has currently approved. A course leaves their library as soon as a new revision opens, and the approval is checked again on every file download. Instructor accounts can open this library and change their own password, and EPHOR Ai refuses them everywhere else, even through a direct link.
Inspectors ask to attend in person or over Microsoft Teams, with a warning if a colleague has already requested the session. The operator confirms with arrangements or chooses a reason it cannot host and explains it, the inspector can ask it to reconsider, and a request still waiting on the operator when the session runs is recorded as not accommodated. Teams joining links are shown only to the observing inspector and the operator, never in email, calendar feeds or exports.
Each change to a session is saved as a new version. A change to the date, time, session type, location, instructor or remote availability sends a confirmed observation back: the inspector reconfirms against the new version, the operator confirms the arrangements again, and both receive a before-and-after summary. A corrected note does not trigger this.
The coverage report, in PDF or CSV, lists every session in the period and counts a session as observed only when an inspector was confirmed and attended. Open requests, requests the operator could not host or never confirmed, and reports still due are counted separately, with totals by inspector and by attendance mode.
Reviewers compare a course folder with any earlier version and see which files were added, changed or removed, and both organizations can comment on individual files. Operators submit PDF, PowerPoint and Word files, PDFs open inline, and EPHOR Ai refuses Office files that contain macros or embedded content.
Inspectors see the published programme in calendar and list views, from simulator, line check and OPC sessions to ground school, CRM, SEP and type rating training, and request to observe straight from it. Times can stay unconfirmed, each session states whether it can be observed remotely, and a CSV import is previewed row by row and cannot be committed until every error is fixed.
Either organization can open a request for the other to decide, or share files for information. Marking a request complete freezes its files as a numbered version, a request sent back must include a note on what needs to change, and the database refuses any request where one organization would both submit and approve.
Observation requests, confirmations and outcomes, approval submissions and decisions, and comments each email the other side once saved, and a failed email never blocks or undoes the record. Messages stay behind sign-in, with email saying only that one is waiting. The operator shares monthly audit PDFs, each with its version history, and both organizations get one PDF summary after each day with programme or approval changes.
EPHOR Ai has no AI of its own. Its strength is what the server and database enforce: which organization may act at each step, which version a decision covers, and a history the database refuses to edit or delete.
Inspectors request observations, the operator confirms them or explains why it cannot host, and only the observing inspector records the outcome. The server checks the organization on each of these actions, and the database refuses an approval request where one side would both submit and decide.
Record events, the audit trail, approval files, comments and certificates are kept in append-only tables that refuse edits and deletions, and the audit trail exports as CSV for any date range. Inside EPHOR Ai, the authority's downloads, report exports and sign-ins are internal events that operator users cannot see.
Password sign-in always requires a six-digit code sent by email, and disabling an account ends its active sessions immediately. A local MCP connector lets a tool-using assistant list and edit training sessions, preview imports and work on approval requests, signed in as a person and limited to that person's permissions.
Agree a representative sample and a baseline from your current process. These are proposed evaluation measures, not published performance results.
A training session change and a document package with a revision.
Inspect the operator and reviewer views, reconfirmation, and approval snapshot.
Approval turnaround; revision queries; reconfirmation completeness.
An inspector requests the session and asks to attend in person, over Microsoft Teams, or either way; Teams is offered only for sessions the operator has marked as observable remotely. The operator can acknowledge the request first, then confirms it with arrangements or chooses a reason it cannot host and explains it. Inspectors can follow their observations in a private calendar feed, where unconfirmed requests appear as tentative. After the session, only the observing inspector records attendance and the outcome.
Each change to a session is saved as a new version. If the date, time, session type, location, instructor or remote availability changes, a confirmed observation goes back for reconfirmation against the new version, and the inspector and the operator receive a before-and-after summary. A corrected note does not trigger reconfirmation. Cancelling a session withdraws its observations and tells the inspectors.
Yes. The operator or the authority can open a request for the other to decide, or share files for information without a decision. The database refuses any request where one organization would both submit and approve. Marking a request complete freezes its files as a numbered version, and a request that is sent back must include a note on what needs to change.
Each approval issues a numbered certificate of approval as a PDF. It lists every approved file with its SHA-256 fingerprint, the version approved, the approver's note and the decision history. The certificate is stored unchanged and emailed identically to the operator and the authority.
No. Instructors see only the version of each course that the authority has currently approved. When the operator opens a new revision, the course leaves the instructor library until it is approved again, and the approval is checked again on every file download.
A session counts as observed only when an inspector was confirmed by the operator and attended. The coverage report lists every session in the period, leaves cancelled sessions out of the percentage, and counts open requests, requests the operator could not host or never confirmed, and reports still due separately. It exports as PDF or CSV.
No. EPHOR Ai has no AI and makes no decisions. Every confirmation, observation outcome, approval and return is recorded by an authorized person, and each approval stays attached to the exact file version it covers.
No. EPHOR Ai is a standalone workspace for training oversight, observation coordination and approval records between an operator and its authority.
As your fleet, routes, rules or teams change, we adapt EPHOR Ai to match, and you approve each change before it goes live.
A focused walkthrough, with the people and decisions that matter to your team.