FAQ - Frequently asked questions

I can. I'm a Johannesburg-based software engineer and data engineering consultant, and I design and build production data platforms for South African businesses: ingestion pipelines, warehouses, dashboards and the machine-learning models on top. For a production example, see the Decarb.Earth case study: one pipeline unifying fifteen inverter ecosystems across 24,350 solar plants.

It depends on scope: a single dashboard over an existing database is a very different project from a multi-source data platform. Two structural choices drive most of the cost: how many data sources need integrating, and whether you pay per-seat licensing in dollars or run open-source tools on fixed-cost hosting in rand. For a concrete answer, use the free scoping tool: nine questions, an indicative cost band in ZAR, no email address required.

Count your dashboard viewers versus builders, and look at the ecosystem you already pay for. Self-hosted Metabase is my default recommendation for cost-sensitive South African teams with their own database; Power BI wins inside Microsoft 365 shops; Grafana belongs in infrastructure monitoring, not business BI. I've written a full comparison: Choosing a BI stack for a South African business.

I'm based in Johannesburg and work remotely with clients across South Africa and internationally. Most engagements run fully remote with regular check-ins; on-site time in Gauteng can be arranged where a whiteboard genuinely helps.

Four stages: Strategy (identify and prioritise the challenges worth solving), Build (rapid prototyping to a working MVP), Scale (hardening into enterprise-grade production systems) and Continuous Improvement (measuring against business metrics and evolving the solution). The full breakdown is on the process page.

Both. Projects move through Strategy, Build and Scale, and the fourth stage, Continuous Improvement, usually runs as a monthly retainer with me as a fractional data lead: scoped in days per month, priced in ZAR, covering platform health, new reporting and a shared improvement backlog. If you'd rather call me only when something changes, ad-hoc work is fine too. The process page describes all four stages.

Systems programming in Zig, C++ and Rust; backends in Go, Elixir with Phoenix LiveView, Node.js and Python; web in TypeScript with React, Next.js, SvelteKit and Astro; data and machine learning in Python and Julia; visualization with deck.gl, Plotly Dash, Recharts, D3 and Power BI; databases including PostgreSQL, MongoDB, MariaDB, SQLite and DuckDB. There's more detail on the about page.

POPIA makes your business accountable for who can access personal information and why. The BI layer is usually the widest door into that data because it's the one tool the whole company logs into. In practice that means designing in row-level access control, choosing hosting regions deliberately, and keeping audit trails of who saw what. I treat POPIA as an engineering requirement from day one rather than a retrofit. (That's engineering practice, not legal advice.)

Both, and in that order: reliable pipelines and modelled data come first, and machine learning earns its keep on top of them. In production I've built a hierarchical pricing model in Julia for Nestify's short-term rental platform and multilingual NLP sentiment analysis for the insurer Hollard.