tumAIni

💡 Inspiration

In Kenya, many women die from cervical cancer not because it's untreatable, but because it's caught too late.

We were deeply moved by the stories of women who couldn’t access early screening, didn’t know their risk, or couldn’t afford care.

We wanted to build a tool that offers more than just diagnostics it offers hope (tumaini in Swahili).

Our goal was to bridge the gap between risk, diagnosis, access, and affordability in a single AI-powered solution.

⚙️ What It Does

Tumaini is an AI-driven cervical cancer risk prediction and care coordination platform.

It enables healthcare workers to:

Input key patient data (age, sexual activity history, HPV/Pap results, etc.)

Get a real-time cervical cancer risk score

Receive personalized screening and follow-up recommendations

Check availability of screening tools (e.g., Pap kits, HPV tests)

Estimate costs and financing options based on insurance status

🛠️ How We Built It

Frontend: React + Tailwind CSS for a responsive, clean interface

Backend: Flask REST API that receives patient data and returns risk predictions

AI Model: Trained on a cleaned dataset using logistic regression and random forest models, with SHAP explainability

Data: Used fields like age, sexual_partners, first_sexual_activity_age, hpv_test_result, smoking_status, and insurance

Visualization: Used Nivo Charts to display screening trends, inventory status, and risk distributions

Deployment: Locally and with simulated data; API ready for cloud deployment

🧩 Challenges We Ran Into

Data Imbalance: Most records were low-risk, so we had to oversample high-risk examples for better model learning

Missing Values: Some fields were incomplete and required careful imputation or exclusion

Modular Integration: Ensuring the AI model, inventory logic, and cost estimation worked smoothly together took extra planning

Designing for Usability: Creating a system clinicians would find intuitive and trustworthy required constant iteration

🏆 Accomplishments That We're Proud Of

We successfully built a working prototype that combines AI, health logic, and finance data

Our system outputs explainable predictions that align with Kenyan clinical guidelines

We created a modular architecture that can be extended to other women’s health challenges

We translated a complex health challenge into an elegant digital experience

📚 What We Learned

The power of explainability in health tech SHAP values helped us build trust into predictions

Cross-domain integration: From AI to supply chain to public health, healthcare tech must connect multiple moving parts

That simplicity in UI/UX matters as much as intelligence in the backend

That working with real (or near-real) health data demands ethical thinking and user empathy

🚀 What’s Next for tumAIni

Connect to live hospital data systems (like EMRs or DHIS2) for real-time use

Add SMS reminders and patient follow-up tracking

Deploy on cloud (e.g., Firebase or Render) for real-world piloting in clinics

Expand to cover other reproductive health conditions like breast cancer and ovarian cysts

Partner with Kenya’s Ministry of Health to bring Tumaini to national screening programs

Built With

• python
• tailwind
• typescript
• xgboost