Founder-led studio · you deal with me directly

One person owns your build. A senior team ships it.

Most studios put a project manager between you and whoever actually writes the code. I don't. I own the work, set the architecture, and stay on the hook for it while a small senior team ships it alongside me. We cover Flutter mobile, Node and TypeScript backends, Postgres, AI integrations with Claude, OpenAI and Gemini, and the occasional bit of hardware. Seven years of this, on products with real users and real money moving through them.

Book a call →See what broke in production ↓
Shahzad Umar
Shahzad, founder & principal engineer
I set the architecture on every build
6,000
drivers in production
30+
products shipped & live
7 yrs
building real software
AI
in production, not just demos

A few of the products we've shipped

The Courier App logoThe Courier AppYDigest logoYDigestCalcount Australia logoCalcount AustraliaWithLyla logoWithLyla

Selected work

What it did. What we built. What broke.

Most portfolios only show the highlight reel. These four are the real version, including what broke once real users showed up, and how we fixed it.

The Courier App

Scaled to 6,000 drivers
Logistics · full-stack
Did
A driver platform for daily deliveries: onboarding, sign-up, and the flows a courier fleet runs on every day.
Built
We inherited a slow app with badly wired flows and code that couldn't scale. We rebuilt the whole thing and took it to 6,000 drivers in production.
Broke
Someone pointed a DDoS at the OTP sign-up endpoint in production, right when new drivers were trying to join.
Fix
We put Turnstile in front of it and stopped the attack without taking the app down. Sign-up got faster, and the data feeding their hiring process got a lot cleaner.

YDigest

500K → 10K API calls
AI · LLM pipeline
Did
A YouTube transcript and summarizer app. You follow your favourite channels and the feed fills up on its own with transcripts and summaries, in whatever format you prefer.
Built
It came to us as a vibe-coded prototype. We rebuilt the data layer underneath it and turned it into something that could actually run at scale.
Broke
Supabase was logging 500,000 REST calls. Runaway requests and a couple of memory leaks were driving the bill up and the stability down.
Fix
We added proper caching, traced the leaks, and got it down to 10,000 calls. Same product, 98% less traffic.

Calcount Australia

70K+ users a year
Mobile · nutrition
Did
A free calorie-tracking app built for a web product that already saw 70,000+ users a year. It covers Australian and New Zealand foods, with barcode scanning built in.
Built
We turned their web product into a mobile app and wired the barcode scanner to a nutrient database built for the region.
Hard part
Nutrition data has to be exact. For someone counting calories, a wrong barcode match or a US figure is worse than showing nothing at all.
Fix
We matched every scan against an AU/NZ food database, so what comes back is the real local number rather than a rough guess.

WithLyla

Flows fixed, fully responsive
Web · rescue
Did
A travel-insurance website for a Hong Kong company, covering the quote, purchase and policy flows travellers depend on.
Built
We came in to rescue it. Every API got reconnected, the flows got wired up the way they should have been, and the caching layer was rebuilt.
Broke
The flows and caching had serious flaws. The site wasn't responsive at all — it fell apart on the phones most travellers were booking from.
Fix
We fixed the flows and caching from end to end, then made it responsive on every screen size we could throw at it.

Hard problems

The parts most studios would rather not touch.

Payments edge cases, device sensors, multi-tenant platforms, store review. This is the kind of complexity that quietly sinks a launch, and we've shipped through all of it before.

SEPA

SEPA & virtual IBANs

Direct-debit mandates, credit-transfer reconciliation, and the webhook lifecycles SEPA actually requires.

PAY

Stripe Connect split-fee

Marketplace payouts to connected accounts, convenience + management fees, subscription paywalls.

RTC

Real-time video (Agora / WebRTC)

Stable join → live → reconnect lifecycle, secure token handling, cross-device reliability.

BLE

BLE & device sensors

Barometric floor detection, hardware pairing, and taming real-world signal noise.

SaaS

Multi-tenant + AI agents

Platform plus client instances, with LLM workflows that hold up in production, not just in a demo.

SHIP

App Store & Play submission

Metadata, permissions and compliance handled so your build passes review without technical rejections.

Flutter app developmentAI app developmentLLM / Claude integrationFull-stack Flutter developerBuild an MVP with AI

How I'd approach your build

How I think, before you've hired us.

The best way to judge a team is to watch how the person leading it works through a problem. So here's how I do.

01

Assess before we rebuild

Inherited a half-built codebase? I'll go through it myself and tell you what's worth keeping and what needs rewriting, with the reasoning behind each call rather than just a verdict.

02

Sequence the risky parts first

Payments, integrations, AI reliability: the things that can sink a launch get tackled first, not discovered the week before you ship.

03

We build with Claude

My team uses Claude every day to ship in weeks what often takes months elsewhere. I still set the architecture and review every critical path before it reaches production.

04

Written, clear, async

You'll always know what's done, what's next, and what's at risk. It's all in writing, so you're never chasing anyone for an update.

What you're actually paying for

We optimise for your outcome, not our billable hours.

If there's a faster or cheaper way to get you the same result, we'll take it and tell you why. That's the opposite of how most hourly shops work, and it's a big part of why clients stay.

Real example
~60 hrs → ~2 hrs

A client asked for a mobile app, roughly sixty hours of work from scratch. We got the same result running on the stack they already had in about two. Right outcome, a fraction of the bill.

Before we start

No surprises, in either direction.

✓

We'll be realistic about timelines

I'll tell you what's in our control and what isn't, like Apple and Play review windows. The dates you get are ones we can actually hit, and we raise risks early instead of at the deadline.

✕

When we're not the right fit

If you need a fifty-person body shop, pure staff augmentation, or someone who'll just nod along, that isn't us. We're a small senior team, and I'd rather say so on the first call than spend your budget proving it.

The studio behind the work

A founder out front, a senior team behind every build.

You always deal with me for strategy, architecture and the hard calls. But you're never depending on a single pair of hands. A small senior team builds alongside me, so we move quickly and nothing grinds to a halt when one person takes a day off.

100%
work shipped by the team, led by me
The Blocship team

How we work

From first call to long after launch.

01

Scope honestly

Before any code, we work out what's actually worth building and what isn't.

02

Architect

We lay foundations that won't need ripping out later: the data model, the integrations, the hard parts first.

03

Ship in weeks

You get working, reviewed software in weeks rather than a six-month black box.

04

Maintain & scale

We stick around afterwards. Real products need a team that still knows them when they grow.