AI engineering

This is not machine-learning research or custom model training. Applied AI engineering means embedding LLM tools into how mobile and web products get scoped, drafted, reviewed, and released — with rules in the repo so output stays consistent.

SolverIQ is a live example: AI roadmaps plus mentor booking on a production platform. This portfolio is Next.js and React built the same way — AI drafts, I review, then it ships. DSSI delivery and booking products use the same discipline on Framework7, Cordova, and PHP.

Read the codebase first → write short enforceable rules → use skills for recurring tasks → prompt with real file references → keep diffs small → review like a senior dev before merge.

Typical deliverables: Cursor or Claude project rules, stack constraints, security bans, review checklists for SQL injection, secrets in repo, weak auth, and contract drift. Teams get templates they can reuse after I leave the engagement.

Published on the blog (June 2026): React Native from zero with rules and skills, a tiny PHP backend skeleton with AI, and AI web developer vs traditional for hiring — all the same fast-drafts, careful-release discipline.

When the product itself uses AI — chat flows, roadmap generation, mentor matching, content suggestions — I work on the integration layer: API calls, prompt boundaries, error handling, rate limits, and keeping user data out of logs.

Demo-quality AI and billable AI are different problems. I scope guardrails, fallbacks, and admin visibility around model output so ops can see what members received.

I lead mobile and full-stack work on live platforms at DSSI (Diverse System Solution Inc.) in the Philippines — food delivery, ride hailing, logistics, booking, e-commerce, and member portals. The flows and stack notes below come from that production work.

dexterbanastao.com is my portfolio and hire-me lane. For the company behind most of this delivery, ride, and commerce software, see dssi.international — context in the DSSI mobile development blog post.

I don't jump straight to screens. Every build starts with a process flow — who does what, in what order, and what status the system shows at each step. That flow becomes the shared map for you, me, backend devs, and QA.

Typical sequence: discovery call → role and feature matrix → process flow sign-off → API contract draft → mobile build per role → integration passes → store or web release → handoff notes. AI speeds drafting at each step; sign-off stays human.

You get visibility at each gate — not a black box and a surprise at launch.

Production mobile for these products is Framework7 plus Cordova — one JavaScript codebase for iOS and Android, native plugins where the product needs camera, GPS, push, or file upload.

I usually split by role: consumer app, merchant or provider app, rider or courier app, and sometimes a lightweight admin mobile surface. Same API layer pattern in each — auth token, role guard, API module, status-driven screens that match the process flow we agreed on.

Framework7 handles routing, lists, forms, modals, and pull-to-refresh patterns that feel native enough for delivery and booking products. Cordova wraps it for the stores. I map each process-flow step to a screen or state so backend devs and QA know exactly what mobile expects at each status.

Before store submission I run device passes on real phones — cold start, background resume, location permissions, and offline or slow-network behavior on the paths that matter for your product.

I'm not siloed on mobile while backend guesses endpoints. I lead with the process flow and a written API contract — roles, status enums, request and response shapes, and error codes the apps already handle.

What I hand backend devs: a status diagram (order/trip/booking/shipment states), JSON samples per endpoint, which role calls which route, and what happens on 401 or invalid transitions. PHP and MySQL on the server side; PDO with prepared statements; thin handlers that match the contract mobile already built against.

During build I review backend PRs for contract drift — field renames, missing statuses, auth gaps — before mobile QA wastes a week. When the contract must change, we version it and update mobile in the same sprint. That coordination is how multi-role platforms stay in sync in production.

On solo or lead full-stack work I own mobile, API shape, and often the PHP layer behind it. AI engineering means Cursor rules in the repo — Framework7 patterns, Cordova plugin usage, API client conventions, security bans — plus skills for recurring tasks like new screens or endpoints.

AI drafts Framework7 pages, API boilerplate, and PHP handlers from the signed-off process flow. I review every diff for security, contract alignment, and store rules before merge. Fellow devs get the same rules and contract doc so AI output stays consistent across the team.

Fast drafts, careful release — whether the product is a delivery app, booking platform, or member portal. See AI engineering expertise and AI engineer services, or the June 2026 blog posts on React Native rules, PHP backends, and AI web developer vs traditional.