Public health researcher. Implementation science, maternal and child health, and the careful use of AI.
Evidence that holds up once it leaves the protocol.
I study how health programs behave in clinics, homes, and districts, and I help organisations decide where AI belongs in that work and where it does not.
A1 · Where programs leak
A program is a chain of handoffs. Watch where it breaks.
One thousand children are weighed. Each handoff that follows loses some of them. The lost do not vanish from the district; they settle into the stratum below, where the register never sees them. Switch an intervention on and watch the stratum thin.
-
Nearly every child is weighed. The chain starts strong.
-
A fifth of weights never reach the register. The form is the problem, not the scale.
-
Counselling happens when there is time and a chart the mother trusts.
-
Referral depends on who is accountable for the handoff, not on the protocol.
-
Distance, cost, and permission decide the last step.
Of 1,000 who begin, 0 reach the end.
B1 · Cross-section
Three strata, one ground.
Research, implementation, and systems are not stages. They are layers of the same ground, and a program stands on all three at once. Open a stratum to read it.
Rigorous, mixed-methods inquiry grounded in real-world settings. I work with questions that matter to communities and to systems: what works, for whom, under which conditions. Evidence is the starting point, but it only becomes useful when it is read in context.
- Formative fieldwork before any protocol is fixed
- Qualitative and quantitative strands designed to answer each other
- Honest reporting of limits, not only results
Moving from evidence to action. I study how interventions are adopted, adapted, and sustained, and how implementation science can make programs more effective and more equitable. Much of my work lives in the gap between what is written into a design and what actually happens in a clinic, home, or community.
- Follow the handoffs, not only the indicators
- Treat fidelity and adaptation as findings, not failures
- Design for the supervisor and the frontline worker, not the dashboard
Health outcomes are shaped by systems of policy, financing, workforce, and community norms. I work with a systems lens to find leverage points and to design interventions that fit the ecology in which they will operate, rather than asking people to bend themselves to the constraints of a program.
- Map who absorbs the consequence of each decision
- Change forms before changing tools
- Build feedback loops that staff already want to read
C1 · Surveyed plots
Three programs, read as research briefs.
Each plot is a traverse through a program. The bars under each station are the share lost there. Each chain broke somewhere interesting.
- Plot 02/1 Maternal and infant nutrition
WINGS Implementation Research Project
Mixed-methods implementation research on how maternal and infant nutrition protocols actually behave inside existing health systems, with growth monitoring treated as an implementation outcome.
Field note. The system rarely fails at a single point. Follow the handoffs, not only the indicators.
- Plot 02/2 Digital parenting readiness
Parenxly Parenting Readiness Platform
Designing a multimodal readiness assessment and evidence-based support pathways for prospective and new parents, with psychometric rigor and clear ethical boundaries around personalization.
Field note. Digital tools work best when they protect nuance, not flatten it.
- Plot 02/3 Surveillance and program design
National Viral Hepatitis Control Initiative
Strengthening surveillance, digital data capture, and behavior change communication so that screening, referral, and treatment could be tracked and improved in near real time.
Field note. Data systems earn trust when they answer the questions staff are already asking.
D1 · On AI
Where AI belongs
Most of the AI I am asked about would be better as a form redesign. Some of it is genuinely new. My advisory work is about telling the two apart before anyone signs a contract: scoping the handoff before the model, keeping data where it was collected, running a tool silently until it has earned the right to interrupt a health worker, and evaluating it the way we would evaluate a drug.
How I work with organisations on AIE1 · Archive
Working notes.
All sheets in the archive- Sept 2026 The 0.9 percent Rapid qualitative analysis appears in less than one percent of implementation science studies. AI may finally make it common. The question is what we trade for the speed. 3 min
- Aug 2026 The red alert and the ASHA A clinical copilot in Nairobi cut diagnostic errors most when it interrupted. What would an interruption mean for a community health worker with no clinical backstop? 3 min
- Jul 2026 Novices with a probe One day of training produced sonographer-level gestational age estimates in Lusaka. The hard question is what happens to the referral chain after the machine says high risk. 3 min
- Jun 2026 Silicon mothers Language models can produce plausible answers about pregnancy in Assam. Plausibility is not validity, and formative research cannot be outsourced to a model that has never missed an antenatal visit. 3 min
- May 2026 ABHA before algorithms India's bet is that identity, registries, and consent come before models. I think the consent manager, not the model, is the real AI policy. 3 min
- Apr 2026 Five thousand alerts, zero outcome studies India's media-scanning surveillance system has flagged thousands of potential outbreaks. Nobody has yet shown it changed the course of one. Surveillance AI needs the discipline we demand of drugs. 3 min
F1 · Correspondence