Skilled Volunteering: Taimaka

06/01/2024

Taimaka is a nonprofit fighting malnutrition in children through highly cost effective treatments. It's only 4 years old, and has treated over 10,000 children at an estimated $1.6k to save a life. 


I led product at Taimaka to make complex patient workflows easier for frontline healthcare teams to understand and improve. I mapped the care process, specified tracking improvements, and explored how ML risk predictions could support clinicians making triage decisions

1. Mapping a Fragmented Patient Journey 

Before proposing features, I interviewed 20+ healthcare staff and mapped what actually happened across admissions, weekly follow-ups, referrals and discharge. The resulting visual system made hidden dependencies, repeated questions and failure points visible to both clinicians and engineers

Problem

  • Limited resources to audit existing product/ prioritize features
  • Little updated user research on top problems experienced by doctors, healthcare workers, and community health workers

Solutions

  • User interviews with doctors, nurses, and  community volunteers
  • Spec for entire mobile flow for easier bug tracking and feature improvements
  • Prioritized list of product improvements/ bug fixes
  • Specs for improved vaccine tracking and twin flows

2. Turning Operational Gaps into Technical Specifications 

Two concrete problems emerged: immunization status was inconsistently tracked after admission, and twin-specific nutrition rules were not reliably applied at follow-up. I translated these into structured engineering specifications:
patient-ID-linked records, age-based vaccination prompts and updates across clinical forms, plus revised twin-tracking logic. 

Results

  • Onboarded 2 additional engineering volunteers
  • Building vaccine tracking and twin rationing programs

3. Communicating ML Model Outputs to Clinicians

I collaborated with ML researchers to implement mortality and nonresponse risk models for triage and how their outputs should appear in clinicians' tools. Beyond model choice, the challenge was designing a useful human decision: which children to flag, what uncertainty to communicate, and how to avoid overwhelming healthcare workers with complex technical jargon. 

Share