Ken Kuria
Systems Engineer &
Data Scientist

Problem

Humanitarian data from sources like HDX is fragmented, inconsistently formatted, and updated on schedules misaligned with decision timelines. Organizations needed a pipeline that could reliably consolidate, validate, and surface analysis-ready data without manual rework each cycle.

Approach

Built an end-to-end pipeline combining Kafka for streaming ingestion, Spark for heavy transformation work, and Airflow for orchestration — ensuring each stage ran in the right order and on schedule. Azure Synapse served as the analysis-ready destination, paired with ETL logic, data validation checks, and reproducible reporting workflows.

Tools & Methods

Scope

Multi-source humanitarian data (Humanitarian Data Exchange). Pipeline designed around reproducibility and governance — holds up as new data sources are added, which matters where the data landscape is rarely static.

Problem

MedInsure sits at the intersection of health data and financial data — two of the most sensitive categories an organization handles. Without a clear governance framework, the platform risked regulatory exposure across jurisdictions and poor product decisions made without solid research behind them.

Approach

Led the data team's work designing a governance and privacy framework drawing on NDPR, GDPR, and HIPAA principles. Conducted healthcare financing and pricing research to ground product decisions in reality, then designed analytics workflows connecting that research to day-to-day feature decisions — giving the build team a consistent, evidence-based foundation.

Tools & Methods

Regulatory Scope

NDPR (Nigeria), GDPR (EU), HIPAA (US) — designed to hold up across all three simultaneously, making the framework usable regardless of which jurisdiction a user or partner is in.

Problem

Teams accumulate large volumes of correspondence and reference material in email, and finding a specific answer means searching manually through old threads. The goal: make that knowledge directly queryable — a person asks in plain language and gets a well-formed answer, not a list of messages to read through.

Approach

Built a chatbot on top of OpenAI's API for question answering and knowledge retrieval over email content. Focused on search relevance — tuning how queries were matched against source content using NLP techniques so the chatbot retrieved the right material before generating a response. Built documentation and testing workflows to track model behavior across query types and catch weak or inaccurate responses pre-deployment.

Tools & Methods

Key Insight

The testing and documentation discipline is the most transferable part — turning an inherently unpredictable component (an LLM) into something whose behavior can be reviewed, compared across versions, and trusted with more confidence. A framework, not just a product.

Expat-Housing

Property listing and housing platform for expatriates in Kenya. Full design and frontend build.

Cadium AI Labs

Full brand and site for Cadium — the agency platform, product pages, and analytics dashboards.

ML Projects

Multinomial Naive Bayes classifier for YouTube spam detection · Authorship style prediction with deep learning · Cervical fracture detection analysis (Kaggle).