Mary Macharia | Software Engineer · Backend & AI/ML
I build backend systems, then teach them to reason.
A software engineer based in Nairobi, working across API design, database architecture, and applied AI/ML. I design, build, test, and deploy systems end to end, and I’m especially interested in the point where a well-built backend meets a genuinely useful AI feature.
A typical request path through one of my backend + AI systems
Backend systems that hold up
REST API design, role-based auth, and database schemas built to isolate data correctly the first time not patched after a bug report.
AI applied to real constraints
Integrating LLM APIs into production-shaped pipelines: structured prompts, output validation, and graceful handling of model failure.
Data as a first-class concern
Modeling schemas, normalizing inconsistent sources, and thinking about data isolation and access boundaries from the start.
Testing and debugging discipline
Comfortable tracing subtle bugs — a missing query filter, a wrong mime type — back to root cause instead of symptom-patching.
Selected work
Featured projects
A mix of backend engineering, applied AI, and full-stack builds case studies below go deep on architecture, decisions, and trade-offs.
Uzima Link
A multilingual, AI-assisted digital health records platform built for Kenyan clinics and community health settings.
Student Progress Tracker
A full-stack FastAPI + Next.js application for tracking individual student progress across a training cohort.
Dataloom
A Data-as-a-Service platform tackling data silos among mid-sized SMEs in East Africa.
Backend engineering
Systems designed to be correct under pressure
From API contracts to database schemas, I design backend systems around the failure modes that actually happen data isolation gaps, inconsistent response shapes, misconfigured access boundaries not just the happy path.
Frontend
Next.js / React client
API Layer
REST endpoints, routing
Auth
JWT, role scoping
Business Logic
Services, validation
Database
PostgreSQL / SQLite via ORM
External Services
AI models, email, storage
AI & Machine Learning
AI that has to work inside a real system, not a demo
I approach AI features the same way I approach backend features: define the problem precisely, validate the data going in, and treat model output as something to verify, not trust blindly. That means structured prompts, schema validation on responses, and clear fallbacks when a model gets it wrong.
See the AI/ML approachHave a backend or AI problem worth solving?
I’m open to backend engineering and applied AI/ML roles, freelance collaborations, and technical conversations.
Get in touch