I build the data pipelines and ML systems that businesses run on.
SQL-first data engineering with dbt and BigQuery, plus production machine learning: demand forecasting, computer vision, route optimization and LLM applications. I am a Data Engineer at Solutech, building for FMCG manufacturers and distributors from Nairobi.
Data and AI Engineer who builds production data pipelines and the machine learning systems that run on top of them.
Currently at Solutech Limited, I own end-to-end analytics engineering — ELT pipelines that power manufacturer-facing dashboards, dbt transformations in BigQuery built on a medallion architecture (bronze → silver → gold), and monitoring of production Prefect pipelines — while leading an end-to-end computer vision project for retail shelf analytics, owning everything from annotation strategy through model training to production deployment. As part of this, I also build customer-facing dashboards in Apache Superset and Looker Studio.
On top of the warehouse I deliver demand forecasting, route optimization and computer vision systems, served as APIs so teams can use them in daily operations.
Outside my day-to-day role, my recent focus has shifted toward agentic AI system design: ReAct-style tool-calling agents, task planning, hybrid memory, and self-reflection loops, packaged as streaming production APIs. I design these systems with cost optimization, auditability, and data-privacy compliance (GDPR-aware handling) built in from the start — the same production discipline I apply to data infrastructure.
Data Engineer
Solutech Limited
Nairobi, Kenya
ELT pipelines, dbt models and BigQuery warehouses that turn raw operational data into trusted, analyst-ready datasets and dashboards in Superset or Looker Studio.
Good fit if your data sits in scattered systems and reports are slow or inconsistent.
Demand forecasting, route optimization and computer vision, deployed as APIs and wired into daily operations.
Good fit if you want predictions or optimization built into how your team works.
LLM applications, retrieval-augmented generation (RAG) and text-to-SQL agents with cost controls and guardrails.
Good fit if your team wants to query data or documents in plain English.
Python ingestion pipelines load raw sales, distribution and retail data from client systems into BigQuery, modeled in dbt across bronze, silver and gold layers and orchestrated with Prefect.
Manufacturers get real-time sales visibility through an embedded Superset layer built on the gold models, with each tenant seeing only their own data.
Forecasting models trained on tested, dbt-modeled sales data and delivered as an ML product on top of the data warehouse.
Route optimization for a distributor with 5,000+ customers and a 200+ vehicle fleet.
A containerized computer vision pipeline measuring Share of Shelf and On-Shelf Availability, fed by dbt-modeled product data.
A modern ELT platform simulating a real production source system, built around the Olist Brazilian e-commerce dataset. An OAuth2-protected FastAPI source service with cursor-based pagination and a background drip-feed feeds a dlt incremental pipeline into a bronze schema, transformed through dbt on a medallion architecture with SCD Type 2 history, all orchestrated end-to-end by Kestra every 30 minutes.
Agentic system that answers plain-English business questions by inspecting the schema, writing SQL, executing it, and reasoning over the result — auto-selecting the right chart or table for the response. Designed with production guardrails: a SELECT-only database role and SQL validator block destructive queries, and Redis caching cuts redundant LLM calls for repeat questions.
Production-ready research agent combining a ReAct agent loop, automatic task decomposition, hybrid (buffer + vector) memory, and a self-critic loop that scores and re-iterates on its own answers. Exposed as an async FastAPI streaming API with real-time progress events, backed by JWT-authenticated multi-user access.
End-to-end MLOps pipeline for customer churn prediction with experiment tracking, data versioning, and automated deployment. Includes model monitoring with Grafana and production deployment on AWS ECS.
Semantic movie recommendation system using sentence embeddings and vector similarity search. Features LanceDB for efficient vector storage and retrieval, with data versioning via DVC and S3.
Intelligent Retrieval-Augmented Generation system for managing GitHub portfolios through natural language queries. Implements query fusion, reciprocal rank fusion, and cross-encoder reranking for high-precision semantic search. Deployed with Docker and AWS ECS Fargate.
Comprehensive sales analytics dashboard connecting Power BI to SQL Server. Features advanced DAX calculations, data modeling, and interactive visualizations for business intelligence.
Comprehensive SQL practice project demonstrating advanced querying techniques including window functions, CTEs, subqueries, and stored procedures on realistic datasets.
Hiring, or have a data or ML project in mind?
Hiring? Download my CV or message me on LinkedIn.
Have a project? Email me what you are trying to achieve, the data you have, and your timeline. I will reply with next steps.