FULL-STACK ENGINEER × AI PRACTITIONER

Work that holds
up to questions.

This is not a gallery of screenshots. It’s a place to try the work, inspect the trade-offs and ask why.

Start exploring

Project workspace

Symply House/ Product platform

verified build

Expo Web build · connected source · source platforms: iPhone · iPad · Android · Web

In-site Device Lab

Device Lab · Desktop Web · connected Expo Web buildOpen full preview

Loading the connected source build…

⌘ Click a project to switchConnected source build · local read-only data · no production effects

Agentic architecture

Capability with
guardrails.

AI accelerates the conversation; contracts, evidence and human review decide what the system is allowed to say or do. The point is not a prompt. It is a measurable engineering system.

01Ground

Resolve the question against the approved corpus and current project context.

sourceIds[]

→

02Decide

Return one bounded answer and, only when valid, one typed action.

GuideAction

→

03Act

Let the UI confirm the project, viewport or evidence state before reporting success.

requestId + ack

→

04Evaluate

Run deterministic fixtures for facts, injection resistance, fallback and follow-ups.

offline evals

Model replaceable Demos remain usable without AI Domain review owns qualityRead the skills

How to read this portfolio

Every claim should
earn trust.

The first build makes the architecture visible: the demos are separate from the guide, the guide is grounded in reviewed content, and uncertainty is shown in the interface.

Truth over polish

Candidate projects stay marked as pending until ownership, source builds and permitted evidence are verified.

Questions become navigation

A guide turn can point to a project, scenario or evidence card without taking control away from the visitor.

Safe by default

Every demo is seeded, local and resettable. No payments, patient records, contacts or production writes.