Forward Deployed Engineer · Johannesburg, South Africa (remote, US hours) I go into a business, map how the work actually flows, and build the system that runs it.
pursuitreadiness.com · john@pursuitreadiness.com · LinkedIn
Most implementations fail not because the technology is hard, but because nobody mapped how the work really happens before they started building. That mapping is the job.
I build systems where the deterministic parts stay deterministic, models are used only where judgement is genuinely needed, and every outbound action passes a named human. The rule I hold on every build: the system can block, it can never send.
Twenty years in sales and marketing before this, most recently as founder of Pursuit Readiness (2023–present). That background is why I build for the person who has to use the thing on a Tuesday afternoon, not for the demo.
ai-portfolio — governance layer for AI systems
A POPIA-compliant data protection framework that runs in production and underpins every system I deploy. Three modules, each enforcing something that would otherwise depend on someone remembering:
audit_trail.py— SHA-256 hash-chained event logging with tamper-evident verification. Every PII operation recorded with UUID, timestamp, actor, resource, action and trace ID.consent.py— scope-based consent registry with capture, persistence and runtime enforcement. PII operations check consent before execution.hitl.py— human-in-the-loop approval gates for data access, deletion, correction, escalation and legal review. Mandatory, and not bypassable programmatically.
Built in 48 hours because a client needed POPIA guarantees before they would deploy. It turned out to be reusable, so it now sits under every build.
Salesforce Trailhead — nine badges, October 2026 · verify all nine Data Modeling · Data Management · Data Quality · Reports & Dashboards for Lightning Experience · Business Rules Engine · Accessibility and AI Foundations · Variables and Field Types · Well-Structured Data · Data Literacy Basics
GitHub — GitHub Copilot, Developer Program Member
Microsoft Azure — Image analysis and generation with Azure AI
- Deterministic before agent. If a rule can decide it, it is code, not a prompt.
- A named human approves every send. The system can block. It cannot send.
- The receipt logs the doing, not the content. Stage, actor, timestamp, request ID.
- Measurement is computed from receipts, never typed by hand.
- I don't list a tool I can't demo live.
Currently open to forward-deployed engineering, solution architecture and AI implementation roles.