RFP for Artificial Intelligence Analytical Engine

Position description

Global Support and Development (GSD) invites proposals from qualified software development firms to design and build a Google Gemini-driven analytical engine and web engagement platform (or 'harness') for the Disaster Preparedness & Response Readiness Index (DPR2I) initiative, utilizing the Google Cloud Platform (GCP).

BACKGROUND & OBJECTIVES
The DPR2I establishes standardized benchmarks for national readiness with targeted partners in the Caribbean region. To maximize operational efficiency, GSD is modernizing the DPR2I architecture to leverage an intelligent, purpose-built engine capable of ingesting source documents, reasoning through alternative data stores, and generating validated capacity assessment reports automatically using programmatic guidelines.

CORE SCOPE & REQUIREMENTS
The vendor will implement a "Rapid Minimum Build Strategy" to deploy a secure, responsive web application delivering six core components:
1. Authorization & Country Scoping (Secure login, strict multi-tenant data partitioning)
2. Upload & Extraction (Drag-and-drop PDF/Word/Excel text and metadata parsing)
3. Run Orchestrator (Modular prompt architecture & AI API execution across functional modules)
4. Calculation Module (Algorithmic execution of DPR2I formulas and index scoring)
5. Report Renderer (Automated generation of downloadable executive reports and scorecards)
6. Dashboard Management & Run State Tracking (Historical tracking across successive runs)

ELIGIBILITY CRITERIA
This Request for Proposals is open solely to established and legally registered business entities (companies, firms, corporations, partnerships, or limited liability companies). Proposals submitted by individual, independent contractors will not be considered.

TIMELINE & BUDGET
- Delivery: Accelerated 4-week rapid build sprint (or vendor-proposed practical alternative)
- Budget target for core build: < $50,000 (excluding hosting/O&M strategy)
- Contract anticipated start date: September 4, 2026

FULL REQUEST FOR PROPOSAL:

REQUEST FOR PROPOSAL (RFP)

Artificial Intelligence Analytical Engine & Engagement Harness for the DPR2I

1. Introduction & Background

Global Support and Development (GSD) invites proposals from qualified software development firms to design and build a Google Gemini-driven analytical engine and web engagement platform (or 'harness') for the Disaster Preparedness & Response Readiness Index (DPR2I) initiative, utilizing the suite of tools available within the Google Cloud Platform (GCP).

****This Request for Proposals is open solely to established and legally registered business entities (companies, firms, corporations, partnerships, or limited liability companies). Proposals submitted by individual, independent contractors will not be considered.

The DPR2I is a tool designed to establish standardized benchmarks for national readiness with targeted partners in the Caribbean region. The DPR2I is built to be complementary to existing regional frameworks, such as the CDEMA Comprehensive Disaster Management (CDM) audits, Country Workplans (CWPs), and the Hyogo/Sendai Frameworks. By leveraging existing data sources rather than creating new data collection expectations, it provides a clear, data-driven "snapshot" of a Partner's current operational standing. Partners are informing the design; it includes common indicators as well as specific ones for their context aligned to their long-term priorities. Ultimately, the model enables the continuous tracking of readiness levels through successive updates that adapt to each Partner’s needs.

To maximize operational efficiency and drastically reduce data-gathering burdens, GSD is modernizing the DPR2I architecture. Rather than relying solely on manual scoring and human resource-intensive data collection, the initiative will leverage an intelligent, purpose-built engine capable of ingesting source documents, reasoning through alternative data stores, and generating validated capacity assessment reports and dashboards automatically using programmatic guidelines and guardrails.

2. Purpose

The purpose of this RFP is to select a technical partner to design and deploy a rapid, secure, and production-ready AI engine and user platform. The tool will serve as the primary web-based interface for national disaster management entities, shifting their level of effort away from tedious manual data-gathering toward high-value quality assurance and strategic validation.

3. Objectives & Context

The primary objective of this assignment is to operationalize an AI engine that automates, standardizes, and streamlines the disaster readiness assessment lifecycle across participating nations. Currently, an initial readiness baseline phase is underway across three (3) participating Caribbean countries, running through October 2026. This phase evaluates readiness against established national, regional, and global disaster management frameworks and external resources, including the CDEMA Comprehensive Disaster Management (CDM) Audit, Sendai Framework, Hyogo Framework, National Disaster Preparedness Baseline Assessments (NDPBA), and country workplans.

Specifically, the AI engine will need to accomplish the following core strategic and technical objectives:

1. Enable Data Search, Synthesis, and Conversion: Automatically search, parse, and ingest source data from both structured and unstructured secondary documents (e.g., policies, post-disaster After Action Reports [AARs], contingency plans, audit documents). The engine will interpret these texts, map evidence directly to indicator rubrics, generate objective scores with justifications and confidence ratings, and feed structured output into the platform’s visual dashboard.
2. Provide Strategic Decision Support: Serve as a broader strategic management tool for partner leadership to identify operational bottlenecks, measure capacity gaps, and prioritize strategic investment priorities.
3. Scalability & Future Expansion: Support phased growth. Following the initial 3-country baseline phase concluding in October 2026, the engine and platform architecture must seamlessly scale to accommodate additional participating countries. The underlying data model, prompt framework, and platform UI must be designed to incorporate new disaster management functions (beyond the initial core modules) or modify existing indicators and rubrics over time without requiring structural software rework or database re-engineering.

4. Intended Users

• National Partners in the Caribbean region but can be expanded to other regions (e.g., National Emergency Management Authorities-NDMOs): Local authorities who upload country-specific documents, trigger automated assessments, and review results for their respective nations.
• GSD and Vendor Administrators: System administrators responsible for managing access, issuing/resetting credentials, and monitoring cross-country assessment pipelines.

5. Scope of Work & Functional Requirements

The vendor will implement a "Rapid Minimum Build Strategy" to deploy a secure, responsive web application focused strictly on core functionalities, avoiding unnecessary user-interface over-engineering.

The application architecture must deliver six core components:

Component

Functional Requirements & Deliverables

Primary Objective Addressed

1. Authorization & Country Scoping

Secure email/password login via a managed identity service. Strict data partitioning is mandatory: each country user account carries a country token ensuring they can never see, access, or reference another nation's data.

Objective 3: Scalability & expansion (enables seamless onboarding of countries post-October 2026 without data spill risks.

2. Upload & Extraction

A simple drag-and-drop file upload workspace supporting PDF, Word, and Excel formats. The system must automatically extract text and capture critical metadata (e.g., title, date, issuing authority) to calculate data confidence scales.

Objective 1: Data Search, Synthesis & conversion (Ingests secondary structured & unstructured data sources)

3. Run Orchestrator

A one-touch "Run Assessment" feature that feeds documents and pre-built prompts to the AI API. It must process individual indicators across specified core modules (currently Maritime Access, Emergency Communications, Emergency Management Coordination, Protection of Vulnerable Populations, and WASH, with more to be added later) while displaying an active progress bar. The orchestrator and prompt architecture must be fully modular, enabling administrators to easily add new disaster management functions or update indicator definitions/rubrics without code modifications or platform downtime.

Objective 1 & 3: Automated Indicator Mapping & Modular Expansion (Executes LLM reasoning across function, easily accepts new model prompts)

4. Calculation Module

Pure algorithmic code to execute pre-defined DPR2I mathematical formulas exactly (including gap calculations, critical-factor doubling, composite readiness indices from 0–100, and currency decay confidence metrics).

Objective 1 & 2: Objective Data Conversion & Decision Support (Converts textual evidence into standardized scores and gap metrics)

5. Report Renderer

Automated compilation of a professionally styled, downloadable report containing a headline scorecard, executive summary, gap-colored tables, per-indicator evidence citations, and a data-provision gap log.

Objective 2: Strategic Decision Support (Translates readiness outputs into actionable policy and investment priority)

6. Dashboard Management & Run State Tracking

The frontend HTML functionality and visual dashboard templates are already fully defined and provided. Integrate and maintain this existing dashboard within the platform across successive assessment runs. The engine must systematically archive and store structured data from all past DPR2I runs per country, enabling the platform to render historical comparisons, generate downloadable reports, and track system-level changes over time.

Objective 2& 3: Systemic Tracking & Phased Growth (Measures longitudinal progress across baseline/endline runs and tracks capacity growth)

6. AI Tool, Platform & Hosting Expectations

• AI Technology Stack (Vendor-Proposed): The AI analytical engine must leverage Google Gemini. Vendors must detail how they will optimize Vertex AI / Gemini API calls, manage prompt engineering, and maximize cost efficiency during large-scale document parsing.
• Hosting & Deployment Infrastructure (Vendor-Proposed): GSD’s enterprise cloud servers and API integrations are standardized on Google Cloud Platform (GCP). The vendor must build, host, and deploy the application within GSD's existing GCP environment. The architecture must natively support secure, low-latency API connections to GSD's surrounding software stack.
• Data Isolation & Security: All file storage, extracted text, metadata, and generated reports must be encrypted at rest and in transit (via HTTPS). Multi-tenant isolation must be strictly enforced database-side using mandatory country-prefix filtering. Country-specific data must remain isolated and cannot be used for public model training.

7. Operations, Maintenance & Long-Term Sustainability Strategy

The vendor must submit an Operations, Maintenance (O&M), and Hosting Strategy outlining a clear plan for long-term sustainability covering at least a 3–5 year horizon. This strategy should assume the vendor will manage O&M and hosting for the first 2-3 years and identify the conditions under which GSD could take on hosting from year 3 or so; indicate the minimum capacity GSD will need to have in place to take over hosting. The strategy must include a costing proposal for the noted time frame.

This strategy must address:

• System Maintenance & Upgrades: Routine software updates, security patching, API model updates, prompt optimization, and technical bug resolution.

• Platform & Data Scalability: Technical architecture plans for smoothly onboarding additional countries and adding new disaster management functional modules/indicator sets beyond the initial 3-country rollout.

• Hosting & Data Governance: Long-term infrastructure monitoring, data backup protocols, cost optimization strategies for cloud hosting and LLM token consumption, and ongoing data isolation guarantees.

• Technical Support: Helpdesk escalation pathways and support structures for GSD and participating NDMO administrators.

8. Management & Operational Arrangement

The selected vendor will operate as an independent technical partner, working directly alongside the core DPR2I program team. Vendors must outline their proposed day-to-day project management approach. This includes defining clear reporting lines, regular status update cadences, and quality assurance workflows that interface with the core DPR2I program team.

9. Required Timeline

The initial platform build is anticipated as an accelerated 4-week sprint to meet immediate program baseline milestones:

• Week 1: Core Engine Development (API prompt orchestration, command-line end-to-end execution, and formula unit-testing).
• Week 2: Web Shell Implementation (Auth deployment, country-scoped permissions, document upload/OCR parsing).
• Week 3: Dashboard & State Management Integration (Wiring existing HTML dashboard components, connecting run orchestrator outputs, configuring historical run storage, and testing report output accuracy).
• Week 4: Refinement & Security (HTTPS enforcement, multi-tenant penetration testing, and an end-user pilot).

Note on Deferred Features: To guarantee this timeline, interactive human validation workflows, advanced AI dashboards, natural language query assistants, and Tableau live connections are deferred to subsequent development phases.

Feasibility Analysis & Alternative Timeline Proposal: GSD recognizes that a 4-week sprint carries delivery risks depending on the vendor's chosen technology stack, QA rigor, and prompt architecture. Bidding vendors must evaluate this proposed 4-week timeline and explicitly state whether they accept it or find it unfeasible.

If the vendor determines that a 4-week build is unfeasible or poses unacceptable technical, security, or quality risks, the vendor MUST:

1. Explain why the 4-week option is unfeasible (e.g., technical bottlenecks, testing constraints, risk of hallucination/error, prompt engineering calibration, security auditing).
2. Propose the most practical, realistic alternative build timeline (e.g., 6–8 weeks) with a detailed breakdown of tasks, milestones, and risk mitigation strategies.

10. Eligibility Criteria

Eligible applicants must:

• This RFP is open solely to legally registered business entities (companies, firms, corporations, partnerships, or limited liability companies). Proposals submitted by individual independent contractors will not be considered.

• Demonstrate a proven track record in developing and deploying production-grade Large Language Model (LLM) applications, custom AI orchestration pipelines, and secure web applications.

• Demonstrate expertise in secure cloud architecture, database-level multi-tenant data isolation, and programmatic text extraction (OCR/parsing) techniques.

• Present a multi-disciplinary team comprising an AI/LLM Architect, a Full-Stack Software Developer, and a Data Security/QA Specialist.

• Show the ability to work effectively under highly accelerated project lifecycles (e.g., 4-week rapid build strategies).

• Provide verifiable proof of enterprise-grade compliance regarding data privacy, showing that proposed models/systems protect sensitive client data from public model training cycles.

11. Evaluation Criteria

Submitted proposals will be assessed based on the following weighted criteria:

• Technical Soundness, AI Architecture, and Security Blueprint (35 points)

• Speed and Implementation Feasibility, and Long-Term Sustainability (20 points)

• Team Qualifications & Technical Composition (20 points)

• Cost Effectiveness and Operational Efficiency (15 points)

• Relevant Past Performance & AI Case Studies (10 points)

12. Proposal Submission Requirements

Applicants are required to submit their proposal split into two separate documents:

Document A: Technical Proposal

1. Organizational Background and Relevant Experience: Documented evidence of developing and deploying production-grade AI/LLM applications.
2. AI Engine & Architecture Justification: A comprehensive explanation of the selected AI tool/LLM, detailing why it was chosen over alternatives, how it handles long-context disaster data, and the methodology behind prompt engineering and management.
3. Data Security & Isolation Blueprint: A rigorous layout of how the application will prevent cross-country data leakage, enforce encryption protocols, and ensure data privacy under enterprise cloud terms.
4. Implementation Workplan: A detailed Gantt chart mapping out tasks across the strict 4-week delivery sprint.
5. Team & Management Composition: CVs of the specific technical staff (e.g., lead AI architect, full-stack developer, security specialist) along with an explicit breakdown of who will manage project delivery day-to-day.
6. 3–5 Year Operations, Maintenance & Sustainability Plan: A detailed strategy outlining long-term hosting, model maintenance, helpdesk support, system updates, and architectural scalability to support expanding country participation and new functional modules post-October 2026.

Document B: Financial Proposal

1. Itemized Budget: A fixed-cost breakdown for the core 4-week build components; GSD estimates this to be less than $50,000 (separate from hosting strategy) but vendors should assess and propose their most competitive budget aligned to the required components.
2. Level of Effort (LOE): A clear table outlining personnel roles, daily or hourly rates, and allocated hours per module.
3. API, Operational & Long-Term Cost Projections: Itemized projections covering (a) variable AI API token usage costs during initial baseline runs, (b) recurring monthly/annual cloud hosting fees, and (c) estimated annual O&M support fees covering a 3–5 year horizon.

Application instructions

Annexes:

1. CVs of key team members
2. Sample of previous related work
3. References or client testimonials

• All proposals must be submitted electronically in PDF format by August 10, 2026 at midnight EST to [email protected].
• Clarification questions may be submitted to the same email address no later than July 31, 2026 at midnight EST. All questions will be compiled and answered in one document and shared in a blind email to all vendors who submitted queries no later than August 4, 2026.
• The anticipated start date for the contract is September 4, 2026.

13. Terms and Conditions

All proposals shall remain valid for a minimum of 90 days from the submission deadline. GSD reserves the right to accept or reject any or all proposals without obligation to provide rationale. The successful vendor must comply with all applicable laws, institutional procurement rules, and confidentiality requirements.

The issuance of this RFP does not constitute a commitment to award a contract. Costs associated with the preparation and submission of proposals are solely the responsibility of the vendor and will not be reimbursed by GSD.