About · Lukas Prinsloo
I'm strongest when the answer isn't already written in the ticket.
I design and maintain systems end to end, lead frontend and QA work, and build agentic engineering tooling for complex multi-repository delivery. The work I do best starts with ambiguous requirements, unfamiliar systems and problems that cross technical boundaries.
The agentic tooling is engineering infrastructure, and I run it as such. I use explicit requirements, automated verification, engineering review and specialised agents to coordinate implementation across repositories while retaining responsibility for architecture, integration and production outcomes.
Selected work
Campaign Creator
Problem: Brands needed to create AR, quiz and survey campaigns without engineering involvement, and the flagship platform those campaigns ran on had to stay operational.
Constraints: A live enterprise system, multi-tenant authentication, geofencing and existing campaigns that could not simply be switched off.
Decisions: Use a strangler-fig approach instead of a high-risk replacement: move one boundary at a time behind serverless, event-driven AWS workflows while the existing platform keeps serving.
What I built: The Vue and Node.js product surface, multi-tenant authentication, AR compilation and delivery, analytics and the migration pieces that let the two systems coexist.
Result: The existing service remained live during the separation. Campaign Creator onboarded a self-serve paying client and was attributed to a major enterprise deal.
GG Anvil
Problem: Turn evidence from competitive gaming, mostly scoreboard screenshots in varying formats, into credible competition records when the extraction is machine-generated and uncertain.
Constraints: Scoreboard formats vary, provider cost, accuracy and latency differ, vision-model output is uncertain and competitive results need an audit path.
Decisions: Treat model output as uncertain input. Test AWS Bedrock and vision-capable provider APIs behind an extraction boundary, then combine extracted fields with confidence and quality signals, deterministic validation and a human-review hold.
What I built: The Discord integration, React portal and serverless vision and OCR extraction pipeline, with a provider abstraction and an audit path from submitted evidence to recorded result.
Result: Uncertain results are withheld for human review instead of being silently accepted, and the provider can change as price and capability evolve without the provider owning the system.
Local transcription
Problem: Build an automatic on-device meeting-to-work pipeline that turns natural meetings into useful work without paid cloud transcription or manual admin.
Constraints: Audio has to stay local. Beyond privacy: speaker diarisation, silence-induced hallucinations, latency and the limits of consumer hardware.
Decisions: Run Faster-Whisper, pyannote and Gemma through Ollama locally with explicit hallucination filters, then let downstream agents consume the transcript files.
What I built: An open-source Python pipeline that runs three concurrent models for transcription, diarisation and analysis, then emits documents, structured notes, action items and tickets or task artifacts for downstream agents and engineering work.
Result: Meetings become usable documentation and tickets without manual delegation and with no third-party transcription fees; runtime processing remains on-device with zero cloud calls, sub-0.5-second latency and under 5GB of VRAM.
QA to frontend to platform
QA taught me to understand failure. Frontend taught me to own what I build. Platform engineering expanded that ownership across system boundaries.
Three roles and two promotions on the same Innereality production platform, each one widening the scope I was responsible for.
Platform Engineer · Jun 2025 to present
I now work across the boundaries of the system instead of owning a single layer. I work directly with the CTO on business logic and technical design: multi-tenant schemas, system boundaries, permissions, blast-radius analysis and deployment safety. I carry work across React, Vue, Node.js, AWS Lambda and infrastructure from requirements through deployment, verification and production support, and I lead and support the frontend developer and QA engineer. I rebuilt the AR experience in React in seven days after a critical AR vendor collapsed shortly before launch. I built agentic engineering tooling for coordinating specialised software agents across complex multi-repository feature delivery.
Frontend Engineer · Jun 2024 to Jun 2025
I moved from proving systems worked to being responsible for building and operating them. I learned an established multi-repository production system quickly, delivered improvements and production fixes across React and Vue applications while servicing live customer campaigns, and took substantial ownership of the operator-facing product: brand customisation, runtime theming, interactive previews and a production forms application. I became the technical bridge to non-technical founders and onboarded the next QA engineer.
Software Quality Assurance Engineer · Jul 2023 to May 2024
I learned how the system behaved and how it failed, then built the infrastructure that proved it worked. I bridged non-technical stakeholders and technical consultants, traced defects end to end, designed and built the core automated test infrastructure from scratch with Python and SeleniumBase, owned UAT and issue triage for production releases, and built an AR game prototype before promotion into frontend engineering.
Twelve years of photography and design before and alongside engineering, a Diploma in Digital Photography, a Software Engineering certificate from Stellenbosch University with HyperionDev, and CompTIA A+ and N+ foundations.
Close mentorship from a CTO with more than 26 years in software has sharpened my systems thinking. I am still building the depth that comes with time, but if I sit with a system for a bit, I can usually smell a bad setup.
Where I'm heading
The ambition is bigger than the current title
This is direction, not a current-role claim. I want to become the go-to platform engineer for multi-tenant systems, lead high-performing small-to-mid engineering teams and mentor engineers in production ownership and failure-path thinking.
I intend to deepen my work on resilient, scalable cloud infrastructure while maintaining hands-on technical depth and growing my open-source contributions.