Map the real process
Find the decision, owner, handoffs, exceptions, evidence and cost of failure before choosing technology.
Dlyra Studios · AI, data & automation
I build systems that make complex work clear.
Reliable AI, automation, forecasting and data products—engineered from the decision backwards, with security, evidence and maintainability built in.
Johannesburg · Open to selected small-business pilots
Trust boundaryLeast privilege
System stateObservable by design
Delivery modeHuman in the loop
RAG & knowledge systemsBrowser automationForecastingReal-time pipelinesInternal applicationsAI infrastructure
Selected systems
Depth you can inspect, without exposing private code.
Independent builds and carefully generalized accounts of implemented systems. Each case preserves the real architecture, workflow and engineering decisions while removing private names, data, interfaces and source code.
View all workEngineering approach
From an unclear problem to an operable system.
Good engineering is not the number of technologies in a diagram. It is making the right trade-offs visible and leaving the system easier to own.
Build the narrow slice
Deliver the smallest end-to-end path that can test value with real constraints and realistic data.
Design for failure
Add authorization, validation, retries, observability, tests and clear human intervention points.
Make ownership clear
Document decisions, train operators, define support and leave a measured path for iteration.
Interactive systems lab
Run a miniature pipeline in the page.
These browser-only simulations make the engineering flow tangible without connecting to a real product, data source or model. Choose a system and watch evidence move through controlled stages.
Input
Governed question
A team member asks how to handle an exception and needs the current approved policy.
Outcome
ReadyRun the simulation to produce a controlled outcome.
Ready to simulate
Small-business consulting
Practical technology for work that is still too manual.
I am building toward a focused independent practice: small, controlled pilots that solve one visible workflow before growing into a larger platform.
Full-stack web applications
Responsive customer portals, internal tools and operational platforms with accessible frontend design, typed APIs and dependable data models.
- React / Next.js interfaces
- Python and Node backends
- Identity, roles and audit
App builders & workflow systems
Low-code experiences that let teams shape forms, processes and focused applications without bypassing engineering controls.
- Versioned schemas
- Approvals and state machines
- Reusable design systems
APIs & business integrations
Narrow, documented interfaces that connect existing tools, eliminate re-keying and keep ownership clear across system boundaries.
- REST, events and webhooks
- BI and reporting connectors
- Idempotency and reconciliation
Responsible data collection
Permission-aware web research and scraping pipelines with source provenance, validation, rate control and change monitoring.
- Public or authorized sources
- Structured extraction
- Freshness and quality checks
AI knowledge & document tools
Search, OCR and grounded assistance across business knowledge with citations, access-aware retrieval and human review.
- RAG and hybrid search
- Document intelligence
- Evaluation and governance
Browser and process automation
Recoverable automation for repetitive portal, spreadsheet, email and back-office work, with evidence and clear intervention points.
- RPA and Playwright
- Scheduling and retries
- Operator handoffs
Quality & security assurance
Authorized frontend, API and workflow testing that finds broken journeys, unsafe assumptions and performance regressions before users do.
- Automated end-to-end tests
- API and authorization checks
- Performance and accessibility
Data products & forecasting
Validated pipelines, operational dashboards and decision-focused forecasts built around lineage, baselines and honest uncertainty.
- SQL and data pipelines
- BI integrations
- Monitoring and drift
About me
AI engineering grounded in mathematical thinking.
I’m Tyron, a Johannesburg-based AI Engineer with a BSc in Mathematical Statistics. I work across the full delivery lifecycle—from problem framing and architecture to implementation, testing, deployment and operational handover.
My interests sit where AI, data and real workflows meet: systems that can explain their evidence, survive imperfect inputs and remain understandable after launch.
BSc Mathematical StatisticsUniversity of Pretoria · 2024
Full-stack deliveryPython · TypeScript · SQL
Production-mindedSecurity · tests · observability
Engineering notes
Going deeper than the project card.
Practical writing on security, scalability, failure modes and the choices that make a system dependable.
Browse all notesStart with the real bottleneck
Have a process that should work better?
Tell me where the work enters, what gets repeated and what a better outcome would look like. We can determine whether the answer is automation, AI, data—or something simpler.
Discuss a process