Abdullah Hanif

Writing

Writing

How I work

I don't have a computer science degree. I also didn't learn by binge-watching tutorials and pasting boilerplates. I learned by shipping a multi-tenant ERP that shops actually use overnight.

No classroom. No copy-paste career.

I don't memorize syntax for sport. I care about infrastructure that stays up: schema, WAL mode, deploy scripts, search under load, and what happens when five cashiers hit the same branch at once.

While a lot of graduates spend years on classroom apps, I independently designed and shipped a production ERP from scratch. The logic didn't come from a lecture — it came from shop floors, failed automations, and fixing things at 2 a.m. I use AI tools heavily for speed. I still own the system design.

Below is the four-step habit I follow before every serious deployment. It's not a slogan. It's how Xero Sphere AI got built.

Process

Four phases before I ship

Traditional coding starts at syntax. I start at roles, failure modes, and cost. Then I pick the runtime, then I stress it on real hardware.

  1. 01

    Map every role before backend code

    Before I write APIs, I split the product into three lanes: customer-facing UI, manager dashboard, and admin control. I walk end-to-end flows for each role — what they click, what data they touch, where the loop can break. I look at real competitor UX, GitHub issues, and how shops actually behave on WhatsApp and POS. If the role map is wrong, everything downstream is expensive.

  2. 02

    Hunt competitor pain — then cut token waste

    I read negative reviews and support complaints on competing tools as if they were my own outages. That becomes the backlog of things I refuse to ship.

    On AI features, the goal is not "call the model on every keystroke." It's zero wasted tokens: AI only on high-impact paths (OCR, weird edge cases), with local backend doing the boring work. You still get an AI-native product feel without burning the cloud bill. On Xero Sphere, heavy fuzzy search stays off the POS hot path for the same reason.

  3. 03

    Pick the stack for load, not for trends

    I'm not loyal to a language brand. I pick what fits the load. For Xero Sphere that meant asynchronous Node.js for I/O-heavy APIs, SQLite in WAL mode with a real cache budget, a custom web frontend, and Android (including NDK) where the phone had to own the WhatsApp path. If a workload needs heavier compute later, I'd reach for C/Rust or Python where it earns its keep — not because it's fashionable.

  4. 04

    Bare-metal deploy, stress it, skip the bloat

    Every serious build gets failure drills: bad inputs, mock API failures, bot-style hammering, and prompt-injection style abuse where AI sits. I don't lean on heavy process managers as a crutch. I deploy to Linux VPS with SSH, tarball releases, and small isolated API restarts.

    Then I measure. Playwright and load scripts against a 1M-product branch on a cheap 1 vCPU / 1GB box. Target I actually hit: cold search p95 around 50ms with five parallel cashiers. Details live in Engineering notes.

Industry proof

Big companies already hire this way

Leaving formal studies wasn't a vibe. It's the same bet Google, Apple, Tesla, and several Pakistani software houses already make.

Global

Google

Leadership has said a degree isn't required. Grow with Google treats strong certificates plus portfolio as degree-equivalent for many roles.

Apple

Tim Cook has said roughly half of Apple's US workforce doesn't hold a four-year degree — they hire for ability.

Tesla

Elon Musk has said exceptional ability matters more than diplomas. Hiring leans on projects and tests.

Pakistan

Devsinc

Coding assessments can open the door without a degree.

Arbisoft

Clear their assessment — they don't demand a certificate as the gate.

Where I fit

I don't apply with a transcript. I apply with Xero Sphere AI — live locations, benchmarks, and a WhatsApp engine you can click.

Read engineering notes Contact