Three Modules, One Workflow — Inside Orbid AI’s Architecture

Why Three Modules, Not One

Most AI tools for tender response try to do everything inside a single model. Upload your documents, type a prompt, get an answer. It sounds elegant, but it fails in medical device procurement because the task is not one task — it is at least three distinct tasks that require different architectures, different data sources, and different validation approaches. Orbid AI separates these concerns into three purpose-built modules: Operator, Arsenal, and Intel. Each module excels at one job. Together, they form a pipeline that turns raw tender documents into submission-ready responses in 46 seconds with 90 percent accuracy.

This architectural decision was not theoretical. It came from the founding team's experience at Comen Medical, an $85 million medical device company where they watched monolithic AI tools hallucinate compliance data, miss regulatory nuances, and produce responses that sounded plausible but would have been disqualified on submission. The three-module architecture exists to solve specific failure modes that single-model approaches cannot address.

Module 1: Operator — The Parse-Match-Comply-Draft Pipeline

Operator is the orchestration engine. When a tender document arrives — whether it is a 200-page PDF from a European hospital group or a structured XML submission from a GPA-aligned procurement portal — Operator handles the first and most critical step: parsing. Parsing means more than extracting text. It means identifying the structure of the tender, recognizing which sections contain mandatory requirements versus evaluation criteria, and mapping individual line items to their corresponding compliance categories.

After parsing, Operator performs matching. Each requirement is matched against the company's product portfolio and compliance documentation. This is where Operator calls on Arsenal and Intel — pulling product specifications from one and regulatory mappings from the other. The matching step identifies which products can satisfy which requirements, and flags gaps where no current product or certificate meets the stated need.

The comply step verifies that every matched claim is substantiated. If a tender requires IEC 60601-1 compliance for an infusion pump, Operator does not simply check that the certificate exists — it verifies that the certificate covers the specific product variant referenced in the response, that the certificate has not expired, and that the testing laboratory is recognized by the procuring authority. This level of verification is what separates a compliant response from one that gets flagged during technical evaluation.

Finally, Operator drafts the response. The draft follows the tender's required format, uses the procuring authority's terminology, and includes cross-references to supporting documentation. The entire parse-match-comply-draft cycle completes in 46 seconds for a standard medical device tender. A skilled human team typically takes 14 days to achieve the same output.

Module 2: Arsenal — The Product Knowledge Base

Arsenal is Orbid's persistent memory for everything related to your products. Every technical specification, every test report, every regulatory certificate, every clinical study, and every installation reference is stored in Arsenal and indexed for rapid retrieval. But Arsenal is more than a document repository. It understands the relationships between documents — knowing, for example, that a biocompatibility test report for a catheter material applies to all products that use that material, or that an updated EMC certificate supersedes the previous version.

Arsenal solves a problem that plagues medical device companies of all sizes: institutional knowledge fragmentation. In most companies, product knowledge lives in the heads of senior engineers, in scattered SharePoint folders, in email threads, and in filing cabinets. When a tender arrives that requires specific performance data for a surgical light, the tender team spends hours — sometimes days — tracking down the right document from the right person. Arsenal eliminates this search time entirely.

The module also handles version control and expiry tracking. When a certificate is updated, Arsenal automatically maps the new version to all products and tenders that referenced the old one. When a certificate is approaching expiry, Arsenal flags it before it becomes a compliance gap in an active tender response. For companies managing 30 or more tenders simultaneously, this automated tracking prevents the kind of certificate expiry oversights that lead to disqualification.

Arsenal grows smarter with every tender. Each response adds new data points, new product-requirement mappings, and new compliance precedents to the knowledge base. After six months of use, Arsenal typically contains enough structured product knowledge to draft first-pass responses to common requirement types without any human input.

Module 3: Intel — The Compliance Knowledge Graph

Intel is the module that makes Orbid AI uniquely capable for international medical device procurement. It maintains a compliance knowledge graph that maps 14 regulatory regimes and their interrelationships. These regimes include EU MDR, FDA 510(k) and PMA pathways, China NMPA, Japan PMDA, South Korea MFDS, Australia TGA, Brazil ANVISA, Health Sciences Authority Singapore, and frameworks under the WTO Government Procurement Agreement.

The knowledge graph is not just a list of regulations. It maps equivalences, mutual recognition agreements, and gap analyses between regimes. Intel knows, for example, that a CE mark under EU MDR does not automatically satisfy HSA requirements in Singapore, but that certain testing standards are recognized by both authorities. It knows which GPA member states accept WHO prequalification as a proxy for national registration, and which require separate in-country registration regardless of international certifications.

For Chinese OEM medical device exporters — the market Orbid was built for — Intel solves the single biggest barrier to international tender success: translating Chinese regulatory evidence into formats that international procurement authorities recognize and trust. A NMPA registration certificate needs to be presented differently for a European tender than for a Southeast Asian one. Intel handles these transformations automatically, ensuring that the compliance evidence in every tender response meets the expectations of the specific procuring authority.

How the Three Modules Pipeline Together

The power of Orbid's architecture is not in any single module — it is in how they work together. When a tender arrives, the flow is sequential but fast. Operator parses the tender and identifies requirements. It queries Arsenal for matching product data and Intel for applicable regulatory frameworks. It then synthesizes this information into a compliant draft response. If gaps are found — a missing certificate, an expired test report, a requirement that no current product can satisfy — Operator flags them with specific remediation recommendations rather than silently omitting them.

This pipeline approach means each module can be updated independently. When a new regulatory regime is added to Intel, all future tenders automatically benefit without any changes to Operator or Arsenal. When a company uploads new product documentation to Arsenal, Operator immediately has access to it for the next tender response. This modularity is what allows Orbid to maintain 90 percent accuracy across diverse tender types and regulatory environments.

The Results of Architectural Discipline

Companies using Orbid AI's three-module architecture report measurable improvements. Response times drop from 14 days to 2 days. Compliance accuracy increases from 60 percent to 90 percent. Win rates improve from 19 percent to 38 percent. One $85 million medical device company processed over 30 tenders in a single quarter — impossible with their previous workflow. These results stem directly from the decision to build three specialized modules rather than one general-purpose tool.

If you are evaluating AI tools for tender response and want to understand how the three-module architecture would work for your product portfolio, schedule a technical walkthrough with the Orbid team.