About
I'm a software engineer building scalable web applications with TypeScript, Next.js, React, and React Native, currently shipping production features at Hilal Software Corporation.
I've led a JavaScript-to-TypeScript migration, authored more than 5,000 unit tests, taken Lighthouse SEO scores from 62 to 100, and pushed accessibility scores past 90: the unglamorous work that makes a product actually reliable.
Recently that's meant AI-powered e-commerce automation and an embeddable chat widget that runs cleanly across Shopify, WordPress, HTML, and React apps without touching the host site's performance.
Work
- 2023
Hilal AI · Conversational Chatbot Platform
6 channelsLead frontend engineerMulti-tenant SaaS for deploying AI chatbots across web, Slack, WhatsApp, Messenger, email, and Shopify from one dashboard: embeddable widget, live agent handoff, knowledge-base training, and analytics.
- Next.js 15
- TypeScript
- SignalR
- Socket.IO
- Shadow DOM
- Embeddable widget (v3) with closed Shadow DOM isolation, held under 150KB gzipped
- Live agent dashboard with SignalR-powered real-time conversation handoff
- Knowledge base training from websites, documents, YouTube transcripts, and Google Drive
- Six-channel deployment: web widget, Slack, WhatsApp, Messenger, email, Shopify
- Analytics suite with sentiment tracking, resolution rates, and agent performance metrics
- Backend-for-Frontend pattern: Next.js 15 proxies a FastAPI backend, adding auth context
- Three parallel real-time protocols — SignalR for agents, Socket.IO for the widget, native WebSocket for system events
- Dual build system: Turbopack for the dashboard, a separate Webpack 5 pipeline for the standalone widget bundle
- Custom OAuth 2.0 / OIDC auth with local JWT verification via jose + cached JWKS
- Every page load paid a 1.5–4s remote JWT validation round trip to an auth server.Local JWT verification in middleware via jose with cached JWKS keys, cutting auth latency to near-zero with the remote call as fallback only.
- One codebase needed both a modern Next.js SPA and a lightweight embeddable widget with completely different build requirements.Separate Webpack configs alongside Next.js's bundler — the generic widget externalizes React from a CDN, the Shopify variant bundles it inline for stores with no React.
- The chat widget had to run on any host site without CSS or DOM conflicts.Closed Shadow DOM isolation, kept under a 150KB gzipped budget via async imports and a Prism-only syntax-highlighting shim.
- Auth latency cut from ~2s to near-zero via a local JWT fast path
- Widget bundle held under 150KB gzipped despite markdown rendering, voice input, and file uploads
- Three-layer permission enforcement across middleware, API routes, and client cache
- 2024
Hilal Software · Amazon Seller Dashboard
91% forecast accuracyFull-stack / frontend leadAI-powered SaaS platform for Amazon sellers: real-time inventory sync, PPC campaign analytics, AI demand forecasting, and an NL2SQL natural-language assistant for querying business data.
- Next.js 15
- TypeScript
- TanStack Query
- shadcn/ui
- OKLCH design system
- Real-time inventory sync with low-stock alerts and multi-warehouse transfers
- AI-powered demand forecasting with reorder-point recommendations
- Amazon PPC campaign manager pulling live data from the Ads API
- NL2SQL AI assistant — plain-English questions answered from live business data
- Embedded Apache Superset BI dashboards, feature-gated by subscription tier
- Next.js 15 App Router on Vercel's serverless edge network
- Two-layer permission model: server-side route guards plus client-side UI gating, both derived from subscription tier
- V3 design system built on OKLCH color tokens, inspired by Mercury's fintech aesthetic
- TanStack Query for server state, Zustand for client state, nuqs for URL-persisted filters
- The auth SDK's middleware added 1.5–4 seconds of latency per request.A local JWT fast path via jose + JWKS, falling back to the SDK only when verification fails.
- 42+ features across 4 subscription tiers and multiple roles made feature gating error-prone.Centralized route-to-feature mapping in one config file that both the sidebar and the server-side gate read from.
- The NL2SQL pipeline could take 10–30 seconds per query, risking user trust.An 8-stage streaming response (thinking → reasoning → intent → SQL → executing → result) that turns the wait into a visible process.
- 1.5–4s latency reduction per request via the JWT fast path
- 42+ features across 7 functional areas, spanning 15+ Amazon marketplaces
- Design system migrated from HSL to OKLCH for perceptually uniform theming
- 2025
Hilal Software & Hilal AI · Marketing Platform
2 brands, 1 codebaseFrontend engineerPublic marketing site unifying both products under one Next.js deployment: dual brand theming, ISR-backed live pricing, dynamic SEO/JSON-LD, and a zero-JS CSS-only mega menu.
- Next.js 15
- React 19
- Tailwind v4
- schema-dts
- SEO
- Route-aware chrome rendering a distinct header, footer, and chat widget per brand
- CSS-only mega menu — radio inputs and peer selectors, zero JavaScript for navigation
- ISR-backed pricing pages with live subscription data and a static fallback
- Auth-aware homepage middleware redirecting logged-in visitors to the app dashboard
- Typed JSON-LD structured data across 8 schema.org types
- Next.js 15 App Router, Server Components by default, client islands only where interactive
- CSS-scoped theme system — Hilal AI's full palette lives under a single .hilal-ai class, zero style leakage
- Decode-only JWT middleware, matched to '/' only, with no signature verification on the marketing site itself
- Two products needed completely different brand identities from a single deployment.A CSS-scoped theme system: the main site's tokens live in globals.css, Hilal AI's complete palette sits under .hilal-ai, ensuring zero style leakage between brands.
- The homepage had to stay crawlable while still redirecting returning authenticated users.A minimal middleware that only decodes — never verifies — the session JWT on '/', delegating real validation to the downstream app.
- Complex mega menus risked bundle size and accessibility regressions.A pure CSS mega menu using hidden radio inputs and peer selectors — zero JavaScript for dropdown interactions.
- Zero TypeScript errors across 40+ components
- 21+ passing tests across 5 test suites
- 10 Architecture Decision Records documenting the build
- 2023
Pebble Work
Team ledTeam leadLed a full-stack team project: collaborative workspace tooling with i18n and motion-driven UI.
- Next.js
- Firebase
- i18next
- Tailwind CSS
- Framer Motion
- Collaborative workspace tooling for distributed teams
- Internationalized UI via i18next
- Motion-driven interactions built with Framer Motion
- 2023
E-Commerce Platform
Full stackFull-stack developmentModern e-commerce storefront and checkout flow built on Next.js, Firebase, and Redux.
- Next.js
- Firebase
- Redux
- Modern storefront and product browsing experience
- End-to-end checkout flow
- Redux-managed cart and application state
Skills
- ReactFrontend
- TypeScriptFrontend
- Next.jsFrontend
- React NativeFrontend
- Tailwind CSSFrontend
- Redux / ZustandFrontend
- Three.js / R3F3D & Motion
- GSAP3D & Motion
- Framer Motion3D & Motion
- Node.jsBackend
- REST APIsBackend
- Unit Testing (Jest)Tooling
- SEO & AccessibilityTooling
- Git / Git FlowTooling
Technologies
- Vercel
- Firebase
- Shopify
- WordPress
- AI Integrations
- i18next
Experience
Software Engineer
Feb 2023 · PresentHilal Software CorporationLed a JavaScript-to-TypeScript migration and authored 5,000+ unit tests; took Lighthouse SEO from 62 to 100 and accessibility past 90+. Built an embeddable chat widget shipped across Shopify, WordPress, HTML, and React, plus AI-powered e-commerce automation, within a cross-functional team of up to 10.
Education
Front-End Development Bootcamp
2023Re:CodedContinuous professional development · self-learning & online coursework
2023 · PresentIndependent
Languages
- ArabicNative
- EnglishProfessional · C2 (EF SET)
- FrenchAdvanced
Certifications
EF SET English Certificate · 73/100, C2 Proficient
2025EF Standard English Test