Full-stack and AI engineer with 2.5+ years in production: multi-tenant SaaS, LLM assistants, Flutter apps and data pipelines. Security lives in the schema, and every merge passes a review gate.
React 19, Next.js and Flutter on one design system, in 7 locales including RTL.
Years shipping production code
REST endpoints on one platform
Programmatic SEO pages built
Languages localised, incl. RTL
I learned to build software without AI first. Now I use it every day, with review gates that make sure speed never costs correctness.
Four flagship builds, then six more systems. Screens show each product with sample data.
A CRM where every company gets its own isolated workspace, covering contacts, deals, leads, invoices, tickets, marketing, telephony and automation.
A Dubai property platform: listing search with an AI assistant that drives the map, a Flutter app for buyers and agents, and an SEO site of about 2,200 pages in 7 languages.
A platform that puts small Indian manufacturers online: each seller gets a branded store on their own subdomain, an app to run orders, shipping and money, and a team behind them that onboards and supports every store.
A marketplace for booking verified photographers and videographers. Customers describe the event, get a suggested package and crew, and pay online; photographers get paid to a wallet.
Businesses link their own WhatsApp number by QR and send personalised broadcasts from a spreadsheet, paced to stay safe, with a full admin console for the operator.
A personal assistant that lives in WhatsApp, chats in the user's language and never lets them forget things: reminders from plain conversation, caring check-ins and a web dashboard on the same thread.
Four fast marketing sites for an agency, a steel foundry and two pump brands, with an AI chat assistant on the edge and a Theme Studio that dresses the sites for 31 Indian festivals on a schedule.
A scraping and enrichment platform across six markets, where deterministic extraction does most of the work and a local LLM only handles edge cases.
An end-to-end pipeline combining Claude, the Higgsfield MCP server and Supabase to produce UGC-style real estate videos for 7 social platforms.
A deterministic AI pipeline that turns rough meeting notes into requirements, decisions, risks and clarifying questions, with persistent project memory.
Each tile is backed by something I've actually shipped.
Tool calling, Google Places grounding and a tag protocol that lets the assistant move the map. OpenAI, Claude, MCP, embeddings.
257 endpoints on one platform. Payments and invoices are idempotent, so a webhook delivered twice is never charged twice.
Row-level security, RBAC and tenant isolation, so another workspace's data can't leak through the UI.
A single-source component library in React 19 and Next.js, localised to 7 locales including right-to-left Arabic.
OTP login, push on APNs and FCM, and a full-screen incoming-call UI with iOS CallKit parity.
Cloudflare Workers, Pages and R2 with Docker and GitLab CI. Branch per task, rebase, atomic commits.
Cipher Craft Private Limited
Digi Trend
Freelance
I use agents to move fast and gates to stay safe. Every change goes through the same five steps.
Requirements become a schema first: tenants, roles and row-level security before any screen.
Work is split across backend, frontend, security and QA agents, each with a narrow brief.
An independent reviewer agent tries to break every change before a human sees it.
A branch per task and a merge request reviewed by people, rebased on main.
Small atomic commits ship through CI. Rolling back is one revert, not an incident.
Short, practical write-ups on problems I've actually had to solve.
