Re-architecting the domain model for veterinary clinical trials.
Building a veterinary-first clinical data platform around the realities of multi-species research, instead of forcing animal studies into systems designed for humans.
A structural market failure: bending human software to fit animal biology.
The clinical research software industry is bias: almost every enterprise Electronic Data Capture (EDC) platform on the market was architected exclusively around a human data model. A human-first architecture implicitly relies on rigid assumptions: one subject, a single consenting person, uniform dosing metrics, and fixed medical dictionaries.
When applied to veterinary research, this structure collapses. Because animals range from individual companion pets to herds, litters, and other complex group structures, data managers are forced to rely on fragmented, high-risk operational workarounds. Every workaround complicates data entry and introduces severe compliance risks.
Deriving the study structure through configuration.
Instead of designing a static interface that forces users to assemble complex studies by hand, I built an architectural paradigm shift. I mapped fourteen distinct veterinary study categories down to five fundamental structural dimensions. In this new architecture, a few declarative answers provided during initial setup dynamically derive, compile, and render the entire platform’s database structure, user views, and validation rule sets.
A data model that adapts
The data model adapts dynamically to the study, shifting from individual animals to group-level metrics or nested relationships based on the species and study structure.
One configuration layer
A centralized, zero-code model that turns study categories into the rules that shape the platform. Schedules, regulatory frameworks, withdrawal requirements, and other study-specific behavior are derived automatically, then can be further customized to fit the needs of each study. All while remaining separate from data-collection forms.
A tailored experience based on roles
Access and interface are driven by role, so each person sees only the fields, tools, and permissions relevant to their work. Irrelevant inputs are filtered out, keeping complex workflows focused and reducing data-collection friction in the field.
The model behind the core.
To prove the system’s resilience, I engineered a highly structured, multidimensional configuration matrix. Here is how a single master control point cleanly branches out to resolve complex multi-species variables automatically.
| Study category | Unit of analysis | Schedule | Endpoint grain | Withdrawal period | Regulatory frame |
|---|---|---|---|---|---|
| Companion animal | Individual | Fixed | Per-animal, per-visit | Show / competition only | VICH GL9 · FDA NADA · EMA |
| Food animal, group | Group / pen | Fixed | Pen aggregate (FCR, mortality, yield) | Slaughter WDP per lot | VICH GL9 · FDA NADA |
| Poultry, laying | Group / pen + environmental | Fixed | Egg rate, quality, hatchability | Slaughter + egg withdrawal | VICH GL9 · FDA NADA · USDA |
| Aquaculture | Tank / net-pen (dynamic) | Event-driven | Tank biomass, sampling, FCR | Harvest WDP | VICH GL9 · FDA CVM · EFSA |
| Reproductive, dam / litter | Dam / litter | Event + sub-schedule | Dam reproductive, litter size | Species-specific | VICH GL9 · FDA NADA |
| Multi-species | Mixed per arm | Mixed per arm | Mixed per arm | Per arm | Per species |
6 of 14 categories shown · 5 structural dimensions. WDP is withdrawal period. The complete matrix outlines fourteen study categories, including Companion Animals, Group Food Animals, Laying Poultry, Aquaculture, and Reproductive Dam / Litter models. Illustrative; abstracted for the concept.
What I’m testing: engineering out human deployment error.
Separating configuration from data collection introduces a powerful defensive design layer for regulatory oversight. In this sandbox environment, I am testing whether automating the data schema setup through five simple configuration inputs can structurally eliminate human deployment error entirely.
Instead of relying on users to configure the database correctly, the platform applies and validates protocol rules before data collection begins. Any resulting change to the study structure, forms, or access is automatically captured by a centralized audit trail, exploring whether fundamental compliance with FDA 21 CFR Part 11, VICH GL9, and EU Annex 11 can be built entirely into how the platform operates by construction.
An experimental platform and evolving concept.
Arken is an active, evolving research project. I continue to treat this sandbox as a testing ground for scaling advanced enterprise features against the realities of animal clinical trials. The architecture was built from day one to support new features without requiring the underlying product structure to be rebuilt each time. The core foundation currently carries and expands upon a growing matrix of complex features, including:
ePRO & eConsent
Unified, tenant-isolated modules for site and field workflows.
Integrated lab diagnostics
Native telemetry feeds and automated IDEXX data ingestion.
Veterinary coding dictionaries
Server-side mapping for VeDDRA adverse-event logging and ATCvet drug classification systems.