Skip to content

hello!

I’mFawas — Senior Full-Stack Engineer & Technical Lead

I build platforms that stay fast under load and scale past 10,000 users at once.

Senior SDE II and team lead at WebCastle Media — full-stack and backend systems for UAE e-commerce, fintech and logistics. Right now I’m building MCMuthoot Capital Loan Assistant; before that I shipped MEMuthoot Exim and YTYateem Opticians.

Ernakulam, India · OPEN TO UAE / GCC RELOCATION

that’s me!
Fawas, standing on a plant-lined deck
ernakulam, between deploys
fuelled by ☕ × ∞

About me

Engineering systems that hold up

(the long & short of it)

01 — The start

Computer Science at the University of Calicut — a B.Sc. at Mamo College, then an M.Sc. at Farook College finished at 3.9/4.0 while already working full time. My first engineering job was building admin portals and departmental apps at Markaz Knowledge City in 2022.

02 — Alongside the code

I joined WebCastle Media as a full-stack developer and spent two years shipping 10+ production platforms for clients across the UAE and Middle East, tuning queries and API response times by up to 45% along the way.

03 — Connecting the dots

Now I lead architecture and delivery for enterprise platforms in e-commerce, healthcare, logistics and fintech, running teams of 8–12 engineers — and increasingly building the AI layer on top: RAG assistants, semantic search, agent orchestration.

what’s in my toolbox

  • TypeScript
  • Node.js
  • Next.js
  • React
  • Express
  • Bun
  • FastAPI
  • PostgreSQL
  • MongoDB
  • Redis
  • Apache Kafka
  • BullMQ
  • Prisma
  • Qdrant
  • Docker
  • AWS

psst — press the button

The memory printer

An old CRT hooked up to a dot-matrix printer. Every press prints a photo from [your camera roll] or a number from something I shipped. Completely useless. Absolutely essential.

nothing printed yet

fawas@crt:~$ ./print --random

ready — press PRINT

$ ▍

VS-84

fawas·Tron

↑ press print

Work

Things I’ve shipped end to end

Enterprise platforms for the UAE and India — e-commerce, fintech, logistics and healthcare. Mostly the kind where the hard part is what happens under load, not what happens on the happy path.

Meghanté product listing, with filters and 61 products
digital gold platform
Meghanté storefront homepage, with the hero campaign banner
~/muthoot-exim
$ git log --format=%s -4
feat(vault): redemption + ERP ledger sync
perf(queue): kafka partitions for txn fan-out
feat(ai): gemini + qdrant support bot
$ bun run build ✓ done in 2.1s
50,000 a day
ME

Muthoot Exim

2025

Buy gold, redeem it from a vault, and keep the ledger honest.

A digital gold transaction platform covering purchases, vault redemptions and ERP ledger synchronisation. Built as a TypeScript monorepo on Bun, Next.js and Express, with Kafka and BullMQ carrying the event load.

  • 50,000+ transactions daily
  • Kafka and BullMQ event queues processing 50,000+ financial transactions a day.
  • An AI support bot on Google Gemini and Qdrant vector search, automating 60% of customer enquiries.
  • ERP ledger synchronisation, so the vault and the books never disagree.
  • Bun
  • Next.js
  • Node.js
  • Prisma
  • MariaDB
  • Kafka
  • BullMQ
  • Qdrant

Client platform · private source

Yateem category grid: sunglasses, eyeglasses, AI glasses and contact lenses
yateem.com/ae-en
Yateem Optician storefront, with the Cartier eyewear campaign
yateem/checkout.config.ts
export const gateways = [
"apple_pay", "google_pay",
"samsung_pay",
"tabby", // BNPL
"tamara", // BNPL
];
my current obsession
YT

Yateem Opticians

2025

Luxury eyewear e-commerce, built for how the UAE actually pays.

Full-stack features for a major UAE optical retailer: product discovery across luxury eyewear, contact lenses and prescription orders, plus the commerce workflows around them — wishlist, store locator, prescription validation and home try-on booking.

  • Apple Pay · Tabby · Tamara
  • Next.js
  • Node.js
  • Tabby
  • Tamara
  • Apple Pay
  • Google Pay
Laundry Hub services grid, each with its own booking action
thelaundryhub.ae
Laundry Hub homepage, with the booking slot picker
~/laundry-hub
$ git log --format=%s | grep ^feat
feat(i18n): full RTL pass, ar + en
feat(dispatch): turf.js zone matching
feat(track): socket.io driver channel
$ assignment latency -40%
right to left, properly
LH

Laundry Hub

2024

Multi-tenant logistics across UAE cities, in Arabic and English.

A multi-tenant logistics platform with fully bilingual interfaces and RTL layout support throughout. Drivers are dispatched geospatially, tracked live, and paid through a regional gateway.

  • 40% faster dispatch
  • Geospatial driver dispatch and delivery-zone management with Turf.js and MongoDB geospatial indexes, cutting assignment latency by 40%.
  • Socket.IO for live driver tracking and status alerts.
  • PayTabs regional payment processing, Redux Toolkit and TanStack Query on the front.

how the RTL actually works

Mirroring a layout is the easy half. The hard half is everything that assumes left-to-right — icon direction, number formatting, date order, carousel gestures, and the dozens of places a hard-coded margin-left quietly breaks the page for half your users.

  • Next.js
  • Express
  • MongoDB
  • Redux Toolkit
  • TanStack Query
  • Socket.IO
  • PayTabs
Talkiyo call history, with per-call billing and the report and block flows
real-time communication
Talkiyo landing page, with the listener and language-picker screens
~/talkiyo
$ k6 run load/calls.js
vus............: 2000
rtc_connect....: p(95)=210ms
wallet_recon...: 100,000+ txns ✓
2,000 at once
TK

Talkiyo

2025

Two thousand simultaneous calls, billed by the minute.

A high-concurrency audio and video platform with pay-per-minute billing on top. The interesting problem was never the calls — it was keeping a double-entry wallet reconciled across more than 100,000 transactions while they were happening.

  • 2,000+ concurrent calls
  • TypeScript
  • Express
  • MongoDB
  • Agora RTC
  • Socket.IO
  • Redis

Client platform · private source

Muthoot Capital about page, describing the group's lending business
RAG fintech assistant
Muthoot Capital homepage, with its loan product entry points
~/loan-assistant
$ uvicorn main:app --reload
INFO qdrant collection ready
INFO pii redactor ✓
INFO injection guard ✓
INFO ragas eval ✓
what I’m on now
MC

Muthoot Capital Loan Assistant

2026 — Present

A loan and KYC assistant that refuses to make things up.

A full-stack retrieval-augmented support platform for loan and KYC queries, streaming answers in real time. Most of the engineering went into the guardrails rather than the generation.

  • RAGAS-evaluated
  • Real-time streaming responses grounded strictly in retrieved documents.
  • PII redaction and prompt-injection guardrails on every inbound message.
  • RAGAS quality evaluation in the loop, so regressions are caught before users see them.

why guardrails, not vibes

A support bot in lending cannot improvise. Every answer is grounded in retrieved policy documents, personal data is redacted before it reaches the model, injected instructions are stripped, and answer quality is scored with RAGAS rather than judged by eye.

  • FastAPI
  • Qdrant
  • OpenAI
  • BullMQ
  • PostgreSQL

Client platform · private source

The lab

Side quests & weekend builds

Where I go to learn something before it shows up in production — mostly AI infrastructure lately.

no tutorials were harmed

Easter egg

You found the fun part.try me!

An old CRT wired to a dot-matrix printer. Every press prints a photo or a number from something I shipped. Completely useless. Absolutely essential.

fawas@portfolio: ~/fortunezsh

➜ ~/fortune ./fortune.sh --random

# real number · muthoot exim50,000+ financial transactions a day▍

runs: 0 · 6 fortunes loaded

Contact

Let’s build something remarkable.

I reply faster than my code compiles ⚡

I am open to UAE and GCC relocation, remote and hybrid roles, and interesting problems generally. Tell me what you are building.