AI Developer to help automate preventive neurology practice operations

1. Background & Objective We are a preventive neurology practice in an active growth phase, seeking to engage a skilled AI developer, or AI consultancy, to design and build an internal clinical report automation tool. This RFP outlines the problem, desired solution, technical constraints, and proposal requirements. The selected developer will be responsible for both the architecture and the build. We are not paying for a standalone scoping engagement — we expect proposals to include a technical approach and pricing for end-to-end delivery. 2. The Problem After each patient encounter, our clinicians manually compile information from multiple disconnected sources (e.g., MRI, multiple blood tests, genetic sequencing tests, Dexa scan, etc.) and produce two clinical documents: a structured PowerPoint report (using our standard template which we review with patients) and a Word document summary with an in-depth personalized treatment protocol for patients (can be from 10-20 pages on average). Specific pain points: • Clinicians spend significant time on post-visit documentation that could be automated • Although there is a template, clinicians must input plan into word doc/PPT format • Inputs (notes, labs, history) arrive from multiple sources in varying formats • No integration exists between the tools currently in use • Manual bottleneck will worsen as the practice scales 3. Desired Solution We want a simple, internal web-based tool (or lightweight application) that enables clinical staff to upload patient inputs and automatically receive filled-in draft reports to review. The tool is for internal use only — it is not patient-facing. Core Workflow The tool should support the following end-to-end workflow: Step 1 — Data Collection & Ingestion A simple interface where clinical staff can upload or input all relevant patient data for a given visit, including: • Clinician / doctor notes (typed or uploaded) • Lab test results from multiple external sources (variable formats — PDFs, structured exports, etc.) • Patient history documents • Any additional clinical inputs relevant to the encounter Step 2 — AI Synthesis An LLM processes and synthesizes all inputs into a structured clinical summary, following a pre-defined format aligned with our reporting standards. The AI output should be consistent, clinically appropriate, and ready for clinician review with minimal editing. Step 3 — Automated Output Generation The synthesized content is automatically populated into two output documents: • A PowerPoint report built on our existing branded template (.pptx) • A Word document summary for MD/NP review (.docx) Both outputs should be polished enough for direct clinical review — not rough drafts requiring significant rework. The clinician's role is to review, edit lightly, and approve (e.g., 80% filled in accurately should be a benchmark to achieve with optional refinement for better accuracy post-build) Step 4 — Secure Storage All inputs and outputs must be stored securely in our existing encrypted Google Drive environment. The tool must integrate with Google Drive as the primary storage layer and maintain HIPAA compliance throughout. 4. Project Scope Phase 1 — Architecture, Build & Delivery The selected developer will be responsible for the full end-to-end delivery of the tool described above. Phase 1 includes: • Discovery: understand our existing tools, input formats, and reporting templates in sufficient depth to build correctly • Architecture design: propose and document the technical stack before build begins • HIPAA compliance implementation: data handling, PHI safeguards, and AI configuration • Tool development: build the ingestion interface, AI synthesis pipeline, and output generation layer • Google Drive integration: inputs and outputs flow through our encrypted Drive environment • Template configuration: populate our existing .pptx and .docx templates with AI-generated content • Testing and iteration: QA against real clinical inputs; refine outputs to meet clinical review standards • Handoff documentation: clear technical documentation for ongoing maintenance • Accuracy aim: 80% accuracy for clinician to optimize further • We are open to the build of a customized open-source solution vs. using an off-the-shelf solution depending on level of effort • May include additional ad hoc optimizations after working with team to improve accuracy of AI solution Phase 2 — Future Scope (Not Part of This RFP) Depending on Phase 1 outcomes, we may explore EHR integration, expanded input types, and additional workflow automation. The Phase 1 developer will be the preferred partner for Phase 2. For Phase 2 we would want to understand how we can learn from this internal tool for a potential future patient-facing solution. In essence, how can we build something from the start to allow for mining critical insights on preventive neurology for the future? Although this will not be in scope in Phase 1, we expect some consideration will be given to this in the initial build. 5. About the Practice We are a leading preventive neurology practice focused on early identification and intervention for neurological conditions. The practice is scaling its operations and building out technology infrastructure with a strong emphasis on AI-enabled care delivery to service the growing demand of worldwide neurological disease. Our practice is based in New York City. This tool is a foundational investment in how we operate at scale. The right developer will become a long-term technology partner as the practice grows.