This is a submission for the Agentic Postgres Challenge with Tiger Data
Breast cancer is the most common cancer among women worldwide. Early detection can drastically improve survival rates, yet delayed mammogram interpretation leaves many women waiting weeks without guidance.
What I Built
MammoAI revolutionizes breast cancer diagnostics by transforming mammograms into clear, evidence-based reports with actionable next steps. The app addresses critical delays in analysis and limited specialist access, providing faster, clinically informed insights directly from medical imaging and literature.
Key Features:
Mammogram Upload & Preprocessing – Users can upload images in common medical formats; the app prepares the image for analysis.
Annotation + Notes – Health specialists can optionally mark areas of interest and add notes to guide the analysis.
Intelligent Diagnosis & Treatment Reports – Fine-tuned models and LLM-powered agents analyze images and notes, querying specialized medical knowledge bases to generate actionable insights.
Comprehensive Final Report – Combines mammogram findings, diagnosis, recommended next steps, and practitioner notes into a single, clear report.
MammoAI achieves the simplest workflow for breast cancer analysis while integrating AI-assisted interpretation and evidence-backed recommendations.
Demo
Testing Instructions
Use the following publicly available mammogram images for testing purposes:
Test Images
How I Used Agentic Postgres
I started by authenticating with the Tiger CLI to get access to manage my services.
Using Claude code, I created a new service through the Tiger MCP. The service spun up smoothly, but the MCP couldn’t import my documents, so I uploaded them manually via the UI.
Once the data was in, I created a fork of the service and filtered it to include only articles I had curated on treatment.
I also filtered the original service with articles I provided on diagnosis. Using the CLI, I confirmed both services were up and running.
In the end, I had two focused services, each tailored to a specific set of insights, ready to explore.
Overall Experience
My overall experience was just fair. I was unable to import the documents I needed using the Tiger MCP, which may have been due to a skill gap on my part. However, I was able to comfortably perform forks and filter the database tables within the MCP.