EKG Instan — AI-Powered ECG Reader

What the company appears to be: EKG Instan is an AI-powered medical diagnostic tool that enables users to send ECG images via WhatsApp or Telegram for automated interpretation by multimodal AI models. The author states it aims to democratize access to cardiac diagnostics in underserved communities, with a focus on emergency medicine and real-time clinical impact.

What changed: The project was submitted as part of the OpenAI 2026 hackathon. It represents an early-stage prototype built using open-source tools and AI providers (Gemini/Qwen), with plans for migration to a GPT frontier model. The author describes significant development work including architecture design, prompt engineering, testing, and deployment.

Single most important open question: Is there evidence of traction or commercial viability beyond the developer's own use cases? The description states no revenue, customers or adoption data are available — only self-reported claims about functionality and potential impact.

The description states that EKG Instan is a medical AI assistant embedded within messaging platforms (Telegram and WhatsApp). Users photograph an ECG strip and send it through these apps. The system analyzes the image using multimodal AI models and delivers a structured 15-section report with confidence scores and pixel-level evidence.

It claims to detect clinical scenarios such as Normal Sinus Rhythm, AFib, STEMI, Poor Quality, or Not ECG, and includes urgency alerts for critical findings.

The author describes it as:

• A bot that operates via webhooks
• An AI-powered diagnostic tool with 68-scenario interpretation capability
• A system designed to work like texting a doctor for consultation

Evidence: Self-reported by the author. No independent verification or demonstration of actual functionality beyond developer claims.

The author positions EKG Instan as an AI assistant that makes cardiac diagnostics accessible without downloading apps or creating new accounts — essentially turning messaging platforms into diagnostic tools.

Key claims include:

• “Send a photo, get a diagnosis.”
• Enables interpretation like a cardiologist
• Works in emergency settings where time is critical
• Designed to reduce delays in ECG interpretation in underserved communities

The project evolved from an idea rooted in real-world healthcare challenges (delayed diagnosis) into a prototype that uses AI and existing communication infrastructure.

Evidence: Self-reported. No external validation or market positioning data provided.

The author states the target audience includes:

• Emergency medicine professionals
• Underserved communities lacking access to cardiologists
• Anyone needing quick ECG interpretation in low-resource settings

They also imply a broader user base of individuals who might send ECGs via WhatsApp or Telegram for consultation.

However, no specific customer segments or personas are defined beyond general healthcare use cases.

Evidence: Self-reported. No evidence of defined customer profiles, segmentation, or market research.

The description mentions:

• A plan-based quota system
• Free trial (2 lifetime scans)
• Paid tiers (10–200 monthly scans)

It does not provide pricing details, revenue models, or monetization strategies beyond the implied subscription structure.

Evidence: Self-reported. No actual pricing data, customer acquisition costs, or financial model described.

The system architecture is described as:

• Node.js + Express + TypeScript backend
• PostgreSQL with Drizzle ORM for persistence
• AI models: Google Gemini (primary), OpenRouter/Qwen (planned upgrade to GPT frontier model)
• Messaging via Telegram Bot API and WhatsApp Business API
• Infrastructure: UpCloud VPS, Docker containers, Caddy reverse proxy, systemd service management

Key technical decisions include:

• Async webhook processing for fast response times
• Runtime model switching without restarts
• In-memory fallback during database unavailability
• Multi-pass verification and uncertainty propagation systems
• Structured clinical reports with confidence scores

Codex was used extensively for development, including architecture design, backend implementation, prompt engineering, testing, debugging, and deployment.

Evidence: Self-reported. No independent technical review or performance metrics provided.

The author claims:

• Successfully analyzed ECGs from test users
• AI interpretation matches cardiologist-level accuracy for common scenarios (Normal, AFib, STEMI detection)
• Production-ready with 30/30 automated tests passing
• Deployed on UpCloud VPS with Docker and systemd service management
• Has a working webhook-based architecture supporting large-scale messaging platforms

However, there is no evidence of:

• Real-world user adoption or customer base
• Revenue generation
• Market traction or commercial partnerships
• Clinical validation beyond internal testing

Evidence: Self-reported. No external data on usage, customers, or commercial success.

The description does not mention any direct competitors or competitive landscape. It focuses solely on the author’s own solution and its unique positioning in underserved markets.

No comparison to existing ECG reading tools, AI health platforms, or telemedicine services is made.

Evidence: Not evidenced. No competitive analysis or market positioning data provided.

Potential risks include:

• Regulatory Compliance: The author notes that regulatory compliance considerations are important but does not describe any steps taken toward FDA/CE marking or other approvals.
• AI Accuracy and Safety: While the system includes safety features like disclaimers and alerts, there is no evidence of clinical trials or peer-reviewed validation.
• Scalability Concerns: The current architecture relies on a single developer (Rizki Fauzi) and limited infrastructure. Scaling beyond prototype level may be challenging without additional resources.
• Dependency on AI Providers: The system currently uses multiple AI providers; reliance on a single GPT frontier model introduces risk if that provider fails or changes.
• User Adoption Barriers: Despite ease-of-use claims, adoption in healthcare settings requires trust, training, and regulatory approval — none of which are evidenced.

Evidence: Inferred from self-reported limitations and general industry knowledge. No concrete data on risks or failures.

1. What specific clinical scenarios have been validated through testing? How many ECGs were analyzed in total?
2. Are there any partnerships with hospitals, clinics, or medical institutions for validation or pilot use?
3. Has the system undergone any form of regulatory review or compliance assessment?
4. What is the current status of the GPT frontier model upgrade? Is it ready for production?
5. How does the system handle edge cases like malformed inputs or image quality issues?
6. What are the plans for monetization beyond the described quota-based model?
7. Are there any legal or ethical concerns around AI-assisted medical diagnosis in the jurisdictions where it might be deployed?
8. What is the timeline and budget for scaling beyond the current prototype?