MOHAMMAD ABU OMAR

The only engineer at a fashion manufacturer that produces collections for 19+ brands. I own all of its software: design, build, testing, releases and day-to-day running.

• Built the company's ERP end to end: 43 modules covering products and bills of materials, purchasing, the production floor (cutting, stitching, sampling, embroidery, QC), stock per brand, invoicing and finance, HR, projects and a client portal. 11,000+ commits.
• Automated the operations loop: client intake over WhatsApp and email, invoicing, manager approval, auto-assigned staff tasks, client review and rejection routing, with management reports built around invoices and service lines.
• Built Imagine, the company's AI image and video studio: one place to generate with Veo, Kling, Seedance, Wan, Nano Banana and Higgsfield, with a director that plans multi-shot films, a price quote and billing for every generation, and spending limits.
• Turned plain garment photos into campaign and catalogue images (AI models, multiple angles, virtual try-on, relighting), so brands get shoot-ready visuals without a studio, model or photographer.
• Made garment accuracy measurable: fabric, stitching, prints and trims stay consistent across every frame, colors are checked against the real garment (deltaE), and automatic quality gates reject and regenerate weak outputs.
• Built the AI video pipeline behind full campaign films: scene analysis with FFmpeg, AI shot planning, keyframes, per-shot clips with quality re-scoring and automatic assembly, with resumable jobs and budget limits.
• Kept the platform stable while shipping daily: 7,700+ automated test files, including 113 Playwright end-to-end suites, plus tutorial videos for each module rendered automatically from the real product with Remotion.
• Shipped OVO Studio, an AI brand book generator (guided brief, AI drafting per field, readiness score, AI wordmarks), cutting brand identity work from weeks to hours.

Designed, built and ran Billix as the only engineer: an AI agent platform with 200+ models from OpenAI, Anthropic, Google, DeepSeek, Groq and others.

• Built the full product: Next.js app, Convex real-time backend, Better Auth, Stripe billing, streaming chat and multi-agent orchestration.
• Built a desktop computer-use agent (Electron with a Rust sidecar) that reads the native UI tree before falling back to screenshots, which made actions more reliable and runs resumable after failures.
• Built an AI workflow builder where users describe an automation in plain language and agents build and run it (a Zapier/n8n-style product driven by prompts).
• Built real-time voice mode (Deepgram speech-to-text, Cartesia and OpenAI text-to-speech, interruptions), multi-step Deep Search research and RAG knowledge retrieval.

• Built AI agents and LLM features into the core workflows of Rivalz's decentralized data platform.
• Built RAG pipelines, embeddings and AI data-processing tools for large datasets, and shipped the full-stack features around them in TypeScript, React and Node.js.

• Built and maintained Raydian's product web apps in React, Next.js, Node.js and TypeScript.
• Built APIs, real-time data pipelines, dashboards and data visualizations, and took part in architecture decisions and code reviews.

• Delivered 15+ web apps for international clients in React, Next.js and Node.js, from scoping to deployment and maintenance.

Storm gives a business owner a team of AI bots that handle the inbox, sales pipeline, invoices and reports on their own cloud computers, which the owner can watch live and take over at any time.

• Bots you can reach like people: a bot can have its own email address on the company's domain and its own phone number. People email it tasks, text it or call it, and the bot can call them back. In-app voice calls (iOS CallKit) have live captions and let teammates join.
• Connections: 1,400+ apps through Composio, several accounts per app, access set per bot, and logins and 2FA keys kept in the macOS Keychain. Bots also work in Slack, Telegram and Discord.
• Teams and memory: a main bot hands work to specialist bots, bots message each other in threads, each workspace is its own team, and memory is kept per workspace.
• Safe and proactive: sending, creating and paying ask for approval first, with undo, an activity log, cost tracking and automated evals. Bots also send morning briefs, follow up and watch screens for you.
• Built as four apps in one codebase: a Next.js 16 and React 19 app; an Electron Mac app (menu bar, quick launcher, notifications you can reply to); an Expo and React Native iPhone app with native Swift modules (CallKit, Live Activities, widgets, Share and Siri shortcuts); and a Fly.io edge service. 70+ API areas and 28 releases so far.

• Built 138 animated React components and 25 full landing-page templates with Three.js, React Three Fiber, GSAP and Framer Motion, each installable with one command through the shadcn CLI.
• Built the registry build pipeline (JSON manifests plus llms.txt catalogs so AI coding tools can use the library) and the docs site on Next.js 15, React 19, Tailwind v4 and fumadocs.
• Built the launch: a WebGL particle coming-soon page (custom shaders, bloom), a waitlist with duplicate and spam protection, a branded confirmation email (Resend) and a 47-second teaser video rendered with Remotion.