Back to work
Healthcare · 2025

Juno

Maternal Health Agent

Doctor-grade pregnancy answers. 94% accuracy, 50K+ studies, reviewed by a medical council.

Visit live product
Patient satisfaction4.8/5
Medical accuracy94%
Documents indexed50K+

Mode

Project overview

We created Juno end to end. The name, brand, product design, and the AI that powers it. The founding team were either mothers or trying — they wanted answers they could actually trust. Juno pulls from 50K+ peer-reviewed medical papers, reads uploaded lab results and clinical records, and tells parents what to pay attention to. A medical council signs off on new research before it ever reaches a user. 94% accuracy on complex health questions.

  • 94% accuracy on complex pregnancy and postpartum health questions
  • Reads uploaded lab results and clinical records, breaks them down clearly
  • 50K+ peer-reviewed studies indexed via Graph RAG for multi-hop medical reasoning

Discovery

·

Phase 1/4

Two things to figure out: what kind of product would a parent actually trust with health questions, and which retrieval approach could reason across 50K+ medical papers without falling apart. We tested standard search, hybrid, and graph-based retrieval. Only Graph RAG could connect facts across multiple studies reliably. We also created the Juno brand from scratch: name, logo, visual identity, tone of voice.

Deliverables

  • Brand identity (name, logo, visual language)
  • Retrieval architecture comparison
  • 50K+ document source audit
  • Accuracy benchmark targets

Tech

AI

OpenAIGeminiElevenLabsLangChain

Backend

PythonAzurePostgreSQL

Document Processing

TesseractDoclingLlamaParse
View more work