Rust Systems Engineer · IT Ops · Nairobi

Brian Gachau Thuo

I build reliable backend systems and local AI tooling for teams in East Africa.

99.9% Uptime target
33+ Public repos
Selected Work

Things I’ve built.

Curated projects. Each solves a real problem I ran into.

Terminal example: assigning a laptop via the API and the resulting JSON response
2024 API

Laptop Inventory API

Problem: IT ops needed to track which laptop is assigned to which staff member without a spreadsheet drifting out of sync. Approach: a Rust/Actix-web API over Postgres, using SQLx instead of an ORM so every query is checked against the schema at compile time, plus atomic transactions so unassigning a user always reverts the laptop's status in the same operation — no orphaned "assigned" records. Also ships a server-rendered HTMX dashboard, so non-technical staff can use it without a SPA build step.

  • Rust
  • Actix-web
  • PostgreSQL
  • SQLx
  • Docker
Terminal example: an IT support question answered from a local document with a page citation
2024 AI Tool

AI Support Chatbot

Problem: teams with sensitive internal docs can't route IT-support questions through a cloud LLM. Approach: a fully local RAG stack — Ollama for inference, ChromaDB for embeddings, FastAPI streaming answers over SSE — so questions are answered from ingested PDFs, Word docs, and Markdown with zero data leaving the machine, and zero per-query API cost. Ships with an interactive setup script so a non-developer can stand it up.

  • Python
  • FastAPI
  • ChromaDB
  • Ollama
  • SSE
Terminal example: a CLI budget breakdown using M-Pesa and SACCO categories with an investment recommendation
2024 AI Product

Kenyan AI Financial Advisor

Problem: generic financial-advice tools ignore how Kenyans actually move money — M-Pesa, Fuliza, SACCOs, Chamas, Treasury Bills. Approach: a Clean Architecture Python app (domain/application/adapters) that swaps in local instruments and PAYE tax rules instead of Western defaults, with security taken seriously for a public-facing advisor — HTML sanitization, rate limiting, and a target of 80%+ test coverage enforced via a Makefile lint/type-check/test pipeline. Clearly scoped as educational, not licensed advice.

  • Python
  • AI/ML
  • FastAPI
  • Finance
Terminal example: generating a review briefing that flags cross-file impact and a security concern
2024 CLI

PR Context Builder

Problem: reviewers waste time re-deriving context a diff alone doesn't show — what else touches this code, is there a migration, is there an obvious security smell. Approach: a Rust CLI that gathers diff, commit history, and repo structure from a local repo or a GitHub PR, runs rule-based detection for migrations/config changes/test gaps, then optionally hands that context to an LLM (OpenAI-compatible or local Ollama) to produce a structured review briefing.

  • Rust
  • CLI
  • Git
Brian Gachau Thuo — Nairobi, Kenya · Rust, Systems, AI tooling
About

Engineer first, operator second.

I’m Brian Gachau Thuo, a systems engineer and IT Ops lead at IPA Kenya, based in Nairobi. I design backend systems and AI-powered tools that stay reliable under real-world constraints. My work is mostly Rust and Python, with a focus on local/edge AI, automation, and cloud-native infrastructure. Outside of work, I maintain open-source projects that solve problems I encounter in production.

Languages

  • Rust — inventory API, CLI tools
  • Python — AI chatbots, financial advisor
  • Go — cloud-native tooling

Frameworks

  • FastAPI — async AI services
  • Actix-web — high-throughput Rust APIs
  • SQLx — compile-time checked queries

Tools

  • Docker — deployment & environments
  • PostgreSQL — structured data
  • ChromaDB / Ollama — local RAG stack