About

I'm a Senior Backend and AI Engineer in Johannesburg with 11+ years designing and shipping production software across startups, consultancies and product teams.

11+
years in production software
Johannesburg
remote preferred, hybrid possible
Now
available

I specialise in Python, FastAPI, Django, PostgreSQL, distributed systems and cloud-native architecture. My work spans technical leadership on transactional platforms, a data pipeline that processed more than 200 million email addresses, and production AI systems built on RAG, agentic workflows, structured extraction and human-in-the-loop review.

I stay hands-on from architecture through deployment and mentoring. Lately that has meant founding and building AI-native products: Pairly, CaseNotes, WDDNG and Fashionly.

What I'm looking for

A Senior Backend Engineer, AI Engineer or hands-on Technical Lead role with real ownership from architecture to production. I'm particularly interested in teams building data-intensive products, workflow automation or applied AI. I do my best work where strong engineering fundamentals, pragmatic leadership, continuous learning and close collaboration between engineering and product are valued.

How I use AI

I use AI daily for research, architecture exploration, code review, debugging, test design, documentation and delivery planning. For larger changes I use scoped agents to inspect codebases, propose plans and check work across API, web and mobile. Humans stay in the loop for product decisions, security-sensitive changes and release approval, and I verify output with tests and the real consumer path.

Education and certifications

  • Bachelor of Economics, University of Namibia
  • AWS Certified Cloud Practitioner (2021)
  • Automation Developer Level III, RoboCorp (2023)

How I work

Principles. Not slogans.

Traceable over clever

Retrieval results and citations stay traceable. An AI answer, or a passing unit test, isn't production proof.

Isolation by default

Tenant data is isolated with row-level security and request-scoped tools.

Humans approve what matters

Consequential operations need explicit approval, backed by audit events, token limits and human review.

Fail safe, not silent

Typed outputs, validation and evaluations constrain what models can do, and failures degrade gracefully.

Rather hear it from me? Call my twin.

See my experience ›