Documentation has always been the weak link in service dog credentialing. A PDF with a trainer's signature proves very little about what actually happened in the field. The Training Plus program, operated through officialservicedog.com, was designed to close that gap by building a technical pipeline where the documentation itself carries verifiable provenance. This article examines how that infrastructure works at the system level.
The focus keyword throughout this piece is Training Plus documentation. Understanding the architecture matters not just to engineers building similar systems but to trainers, handlers and ADA compliance specialists who need to know what "verified" actually means when that word appears on a credentialing record.
What Training Plus Actually Requires
The Training Plus standard is built around task-specific evidence. A trainer cannot simply attest that a dog is trained. The program requires video demonstrations of each qualifying task, structured session logs tied to those videos and a trainer identity chain that connects the submission to a licensed or credentialed professional.
Tasks are mapped to disability categories recognized under the Americans with Disabilities Act and aligned with IAADP minimum standards. Each task submission must show the dog performing on cue, with the handler present in frame, in a real-world or simulated public access environment. This requirement is not incidental. It exists because public access behavior under the ADA standard is distinct from obedience performance in a controlled training yard.
The documentation standard was also designed with the DOJ's two-question rule in mind. Under current federal guidance, businesses may only ask two questions of a service dog handler. The Training Plus record is structured so that a compliance officer reviewing it can quickly confirm task specificity and handler-dog team identity without requiring the handler to disclose their disability diagnosis.
The Trainer Portal Architecture
The trainer portal is the intake layer of the entire system. It is a role-authenticated web application that grants access only to trainers who have completed identity verification and credentialing review. A trainer account is not created by self-registration alone. The credentialing step involves document review by the clinical and training oversight team at TheraPetic® Solutions Inc., confirming professional standing before any submission capability is unlocked.
Once authenticated, the portal presents trainers with a structured intake workflow. Each submission is organized around a specific handler-dog team record. The trainer cannot submit a generic video pool and assign footage later. The architecture enforces a one-to-one relationship between a piece of evidence and a specific task claim for a specific team at the time of upload.
Session logs are entered through a structured form, not a free-text field. Fields include task category, behavioral criterion, environmental context, pass or remediation outcome and trainer notes. This structured input matters for downstream ML processing. Free-text fields produce unstructured data that cannot be reliably scored or flagged by automated review modules. Structured form data feeds directly into the validation pipeline.
The portal also enforces file type and resolution constraints at the client layer before upload begins. Video submissions must meet minimum resolution and frame rate requirements so that later computer vision analysis can operate reliably. A submission that fails these technical constraints is rejected at intake with a specific error code, not silently accepted and then failed downstream.
Upload Validation and Metadata Extraction
When a video file clears the client-side pre-check, it enters a server-side validation pipeline. The first stage is format and codec verification. The system checks container format, video codec, audio track presence and bitrate against acceptable ranges. Files that pass this stage move to metadata extraction.
Metadata extraction is where provenance work begins. The system reads EXIF and XMP metadata embedded in the file by the capture device. For video files originating from modern smartphones and body cameras, this metadata includes original capture timestamp, GPS coordinates if location services were active, device make and model and in some cases lens and sensor identifiers.
The extracted metadata is stored separately from the video file itself. This separation is intentional. If a file is re-encoded or transcoded at any point in the pipeline, the embedded metadata may be altered. By extracting and hashing the metadata at intake, the system preserves a point-in-time record that can be compared against the file state at any later audit.
The system also computes a cryptographic hash of the raw file at intake. This hash is stored in the submission record and re-verified at each processing stage. Any modification to the file, whether intentional manipulation or accidental corruption, produces a hash mismatch that triggers a hold and flags the record for manual review.
Timestamp and Geotag Verification
Timestamp and geotag verification is the component that separates Training Plus documentation from a simple video upload. The goal is to confirm that the training evidence was captured when and where the trainer claims it was captured.
The verification logic checks three things. First, it compares the file's embedded capture timestamp against the session log date and time entered by the trainer. A mismatch beyond a configurable tolerance window flags the submission for human review rather than auto-passing it. Second, if GPS coordinates are present in the file metadata, the system checks whether the coordinates are plausible for the claimed training environment. A submission logged as a public access assessment in a retail environment but with GPS coordinates placing the recording inside a private residence will flag.
Third, the system checks for timestamp anomalies that suggest file manipulation. Clock rollback, creation date that precedes capture date and timezone inconsistencies between device metadata and claimed location are all flagged. These checks are not foolproof. A sophisticated actor could manipulate metadata at the device level before upload. The system is designed to catch inadvertent inconsistencies and low-sophistication manipulation, not to function as forensic-grade tamper detection for adversarial cases.
For trainers operating in areas with unreliable GPS signal, the system provides an alternative verification path. A trainer can initiate a session from within the portal's mobile companion app, which logs a server-side session token with a trusted timestamp at session start. That token is embedded in the session record and matched against the submitted video at upload time. This path does not rely on device GPS but still provides temporal verification anchored to a trusted server clock.
The Reviewer Workflow Pipeline
Not every submission requires human review. The pipeline uses a tiered triage model. Submissions that pass all automated validation checks and fall within normal parameter ranges for their task category move into an expedited queue. Submissions that trigger any flag from the metadata checks or format validation move into a human review queue with the specific flags surfaced to the reviewer.
Human reviewers on the Training Plus oversight team work through a review interface that presents the video alongside the structured session log, the extracted metadata summary and any flags from the automated pipeline. The reviewer does not need to hunt for inconsistencies. The interface surfaces them. The reviewer's role is judgment, not discovery.
Each reviewer action is logged with a reviewer identifier, a timestamp and a decision code. The decision codes are not binary pass or fail. They include categories for "pass pending additional session," "trainer consultation required," "technical remediation required" and "escalate to clinical oversight." This granularity is important because it allows the program to support trainers through remediation rather than simply rejecting submissions and leaving the trainer without guidance.
The reviewer workflow also enforces separation of concerns. A reviewer who has a credentialing relationship with a specific trainer is flagged out of that submission queue to prevent conflicts of interest. This is a basic governance control, but it is one that many credentialing programs operating without a formal technical infrastructure do not consistently enforce.
For submissions that reach the clinical oversight escalation level, the TheraPetic® Solutions clinical team, which includes licensed clinical professionals, reviews the record with attention to whether the documented tasks are appropriately matched to the handler's disability category. This review does not require the handler to disclose a diagnosis. It is a task-appropriateness check based on the task category fields already present in the structured submission form.
Documentation Integrity from Submission to Record
The final stage of the pipeline is record generation. When a submission completes the review workflow and receives a passing determination, the system generates a permanent documentation record for the handler-dog team. This record is an aggregated, immutable summary of the verified submission data.
The record includes the verified task list, the trainer identity credential chain, the session dates and locations as verified by the metadata pipeline and the reviewer determination. The record does not include the video files themselves. Video files are retained in secure storage and are available for audit but are not distributed as part of the handler-facing record. This design choice reflects both privacy considerations and practical data size constraints for a credential that may be presented via mobile device.
The handler-facing credential links to a verification endpoint that returns the current record state. If a record is placed on hold after issuance due to a post-issuance audit flag, the verification endpoint reflects that state change immediately. The credential itself does not need to be reissued or recalled. This is a meaningful advantage over paper-based or PDF-based credentials, which have no live state.
Record integrity is maintained through the same hash chain established at intake. Each stage of the pipeline that modifies or annotates the record logs a new hash that includes the previous stage hash as an input. This produces a linear audit trail where any unauthorized modification to an intermediate stage would break the chain at that point and be detectable.
What This Means for the Field
The Training Plus technical infrastructure is not the most sophisticated system imaginable for service dog documentation. It is the most operationally practical system that can be adopted by working trainers using standard smartphones and existing training workflows.
That constraint is a design choice, not a limitation. The goal of the program is not to build a laboratory-grade provenance system. The goal is to raise the floor on documentation quality across a field where the floor has historically been very low. A trainer operating under the Training Plus standard produces documentation that is timestamped, geotagged, task-specific, reviewer-approved and cryptographically hashed. That is materially more trustworthy than an unsigned PDF or a letter on letterhead.
For AI and ML engineers building on top of this infrastructure, the structured data layer is the most valuable asset. Verified task-performance videos paired with structured metadata and expert reviewer determinations constitute a labeled dataset with genuine provenance. That kind of dataset is rare in the service dog domain. The TheraPetic® AI platform leverages this data structure for training computer vision models that can assist with automated PAT (Public Access Test) assessment and gait analysis for task performance evaluation.
For ADA compliance specialists and disability advocates, the infrastructure answers a question that the ADA itself does not: what does a trustworthy training record actually look like? Under current federal law, no documentation is required of service dog handlers. But when handlers choose to carry documentation, the quality of that documentation shapes how they are treated in practice. The Training Plus standard gives handlers documentation that can withstand scrutiny without requiring them to disclose more than the law requires.
The program continues to develop in partnership with trainers, handlers and the broader service dog community. Organizations like IAADP and Assistance Dogs International have established minimum training standards that inform the task taxonomy used in the submission forms. The technical infrastructure exists to serve those standards, not to replace the professional judgment of skilled trainers.
For trainers interested in joining the Training Plus program or reviewing current submission requirements, the program documentation is maintained at officialservicedog.com. For verification of existing Training Plus records, the verification endpoint is accessible through officialserviceanimal.com.