You will get an AI Clinical Decision Support Platform built by an Experienced ICU Nurse
Project details
I built Aegis twice.
The first version is a Next.js clinical AI platform. It ingests unstructured medical records, runs them through a multi-model waterfall - Gemini primary, Gemma fallback - evaluates them against clinical criteria, and grounds every output in live PubMed evidence via RAG and pgvector semantic search. Every decision is logged for audit. The fallback solves two real problems: reliability if the primary fails, and rate limit continuity so the system never goes dark.
Then I rebuilt the entire thing in Ruby on Rails over a weekend to benchmark both architectures. The results were striking — meaningfully faster execution and a fraction of the JavaScript payload. Both versions are live - open them side by side and run the benchmarks yourself.
Before any of this, I spent ten years as an ICU and emergency room nurse across fifteen facilities. I have used Epic and Cerner where a slow or wrong answer is not a UX problem - it is a patient safety problem. I did not read about clinical workflows. I lived them.
I will not build you a demo that breaks under real load. I will ask questions most developers never think to ask.
The first version is a Next.js clinical AI platform. It ingests unstructured medical records, runs them through a multi-model waterfall - Gemini primary, Gemma fallback - evaluates them against clinical criteria, and grounds every output in live PubMed evidence via RAG and pgvector semantic search. Every decision is logged for audit. The fallback solves two real problems: reliability if the primary fails, and rate limit continuity so the system never goes dark.
Then I rebuilt the entire thing in Ruby on Rails over a weekend to benchmark both architectures. The results were striking — meaningfully faster execution and a fraction of the JavaScript payload. Both versions are live - open them side by side and run the benchmarks yourself.
Before any of this, I spent ten years as an ICU and emergency room nurse across fifteen facilities. I have used Epic and Cerner where a slow or wrong answer is not a UX problem - it is a patient safety problem. I did not read about clinical workflows. I lived them.
I will not build you a demo that breaks under real load. I will ask questions most developers never think to ask.
Programming Languages
HTML & CSS, JavaScript, TypeScriptCoding Expertise
Cross Browser & Device Compatibility, Performance Optimization, DesignWhat's included
| Service Tiers |
Starter
$3,500
|
Standard
$6,000
|
Advanced
$9,000
|
|---|---|---|---|
| Delivery Time | 14 days | 21 days | 30 days |
Number of Revisions | 2 | 3 | 5 |
Number of Pages | 5 | 10 | 20 |
Design Customization | |||
Content Upload | |||
Responsive Design | |||
Source Code |
Optional add-ons
You can add these on the next page.
Fast Delivery
+$1,500 - $4,500
Additional Revision
+$75
Additional Page
(+ 1 Day)
+$60
API Integration
(+ 2 Days)
+$150
AI Feature Integration
(+ 2 Days)
+$250
AI Model Performance Benchmarking
(+ 2 Days)
+$350Frequently asked questions
About Giannis
Full-Stack Developer | React, Next.js, Node.js | FHIR
Athens, Greece - 9:08 am local time
Full-stack development of PoCs and production-ready applications.
What you get is a developer who has shipped production-grade applications and will treat your $400 job with the same care as a $40,000 one.
Recent work:
Aegis (aegis-ai-cds.vercel.app) - A clinical AI decision support platform I built independently. Multi-model AI pipeline with automatic failover when one model hits its rate ceiling, RAG via Supabase pgvector for semantic search, live API evidence grounding, Zod validation on every boundary, structured audit logging. Built the Next.js version first, then rebuilt the entire backend in Ruby on Rails over a weekend to benchmark both architectures.
FHIR Pulse (fhir-dashboard-five.vercel.app) - Real-time data dashboard pulling live API data through a Node.js Express proxy on AWS with server-side transformation, reducing client payload by 30%.
Summarist (yiannis-summarist.vercel.app) - Full-stack SaaS with Firebase Auth, session management, Stripe-style subscription gating, and tier-based feature access.
What I do well:
React, Next.js, TypeScript, Node.js, Express
REST API integration (Google, Meta, Stripe, Firebase, Supabase, OpenAI, Anthropic Claude, Gemini)
AI integration: chatbots, RAG pipelines, LLM orchestration, prompt engineering
Supabase, Firebase, PostgreSQL
Rapid prototyping and MVP delivery
Production bug fixes and code reviews
Background: Ten years as an ICU and ER nurse before engineering, which means I understand what it costs when software fails under pressure. Dual US and Greek citizen operating through a US LLC, no sponsorship or contractor paperwork needed. Based in Athens, Greece, fully available during US Eastern business hours for async and live collaboration.
Available immediately. I respond to messages within a few hours during weekdays.
Steps for completing your project
After purchasing the project, send requirements so Giannis can start the project.
Delivery time starts when Giannis receives requirements from you.
Giannis works on your project following the steps below.
Revisions may occur after the delivery date.
FHIR connection and data mapping
I establish the connection to your FHIR server, map the relevant resources, and confirm what data is available before building anything.
Core dashboard build
Patient list, vitals display, and primary clinical views built and deployed to a staging environment for your review.