Companion Healt Pro

Companion Health Pro is an AI copilot that transforms complex medical histories into clear insights, empowering users to understand and take control of their health

Comment 2

Inspiration

Human health is not an isolated event; it is a continuous trajectory of data. However, today's patients face a highly fragmented ecosystem: medical reports trapped in PDFs, blood tests on paper, and diagnoses scattered across multiple hospital portals. This disconnection generates immense anxiety and prevents patients from identifying critical, long-term health trends. Companion Health Pro was born from a profound conviction: the rightful owner of medical data must be the patient, and that data must be understandable, visual, and actionable.

What it does

Companion Health Pro acts as an advanced, AI-powered health copilot. It allows users to upload any historical medical document—from a blurry smartphone photo of an old lab result to complex digital clinical reports. The platform standardizes this information, normalizes health metrics across different laboratory ranges, and translates complex medical jargon into clear, visual trends and actionable insights, putting health control back into the users' hands.

How we built it

We maximized the Google Cloud ecosystem to create a robust, secure, and production-ready AI-native architecture:

  • Google Cloud Vertex AI SDK (@google-cloud/vertexai): Migrated from standard public libraries to the enterprise SDK to guarantee low latency, premium quota management, and rigorous data-isolated environments.
  • Gemini 2.5 Pro (gemini-2.5-pro): Deployed as our core reasoning engine. We exploit its massive 2-million token context window to ingest years of multi-format clinical histories simultaneously without context degradation.
  • Google Antigravity Framework: Powers our autonomous background orchestration. Autonomous agents manage data ingest streams, running semantic verification loops across clinical documents to optimize query accuracy.

The Mathematical Heart & Normalization

To deliver true predictive trend lines despite varying laboratory reference ranges, we implemented a historical biomarker normalization algorithm. We calculate the health insight index \( H_{\text{insight}} \) as follows:

$$H_{\text{insight}} = \sum_{i=1}^{n} \left( w_i \cdot \frac{x_i - \mu_i}{\sigma_i} \right) + \int_{t_0}^{T} f(t) \, dt$$

Where \( x_i \) represents the specific biomarker value, \( \mu_i \) and \( \sigma_i \) are the mean and standard deviation of that specific lab, \( w_i \) is the clinical weight assigned to the marker, and the integral calculates the cumulative trend of the health function \( f(t) \) over time \( T \).

Challenges we ran into

The ultimate technical and ethical challenge was data privacy. Working with sensitive medical information required a zero-trust architecture. We designed a secure pipeline driven by Antigravity agents that automatically anonymize and strip all Personally Identifiable Information (PII) at the server layer before any file payload is processed by the LLM, ensuring strict compliance with healthcare security standards.

Accomplishments that we're proud of

We successfully moved from a passive "data viewing app" to a proactive "AI-Native health agent." We are incredibly proud of our multimodal ingestion pipeline, which can accurately interpret poorly lit document photos and convert them into structured, mathematically normalized data lines that provide real value to both patients and their doctors.

What we learned

We learned that the true power of AI in healthcare does not lie in replacing clinical judgment, but in empowering the patient. By translating the fear of an unknown diagnosis into clear, visual data, we can transform passive worry into active prevention and meaningful conversations with healthcare professionals.

What's next for Companion Health Pro

Over the next 90 days, we will focus on:

  1. Refining our autonomous agents for a closed beta phase.
  2. Expanding multi-language medical terminology support.
  3. Securing end-to-end HIPAA-ready cloud compliance to prepare Companion Health Pro for a global MVP launch.

Built With

Share this project:

Updates