Saviour Ise

As CTO I put Django and PostgreSQL behind it rather than a JavaScript backend, because lead enrichment and email verification are long-running jobs and I wanted them as ordinary background workers instead of something bolted onto a request cycle. Redis for the queue rather than a managed service, since the product was pre-revenue and infrastructure cost had to stay near zero. The constraint was three months to something demoable, built across the whole stack.

The decision that shaped everything else was Flutter for the mobile apps: one codebase serving both stores instead of staffing two native tracks, which a team this size could not have carried. I kept the admin panel and landing page on React and Next.js so the web work could ship on its own schedule rather than waiting on app releases. The constraint was three separate products with one team.

The call I made as CTO was about scope, not stack. An AI productivity platform can expand forever, so I cut the first release down to the agent surface that could actually be demonstrated and deferred the deeper analytics until there was usage to analyse. The constraint was three months to a platform people could be shown rather than told about.

I decided the report was the product, not the dashboard. Anyone can render a chart; the thing being replaced was hours of manual spreadsheet work, so the export path got built first and the interface was designed around producing it. The constraint was real consumption data, which arrives messy and inconsistent rather than clean.

The frontend decision that mattered was treating the child profile as a single component reused across all three surfaces, so a profile renders identically whether a visitor is browsing, a sponsor is reviewing, or an admin is editing. That kept the data shape honest and stopped three versions of the same thing drifting apart. The constraint was AngularJS, an older framework, across a nine month build.

On a payments product the failure states are the product, so I built them as real interface rather than error toasts. A failed airtime purchase where money has already left the wallet is the case that decides whether anyone trusts the thing, and it deserved as much design as the happy path. The constraint was AngularJS across landing page, user dashboard and admin.

As CTO I decided against a CMS. An agency updates its own site a handful of times a year, and a CMS would have added a permanent maintenance cost for that. A static React build meant nothing to run and nothing to patch. As the frontend developer the job was then typography and restraint, because on an agency site the site itself has to be the argument for hiring them.

Working solo across the whole stack in a month, the decision was to own no infrastructure. Firebase holds the content so the studio can change their own work without me, and Cloudflare serves it, which meant there was no server to provision, secure or maintain. Anything requiring ops would not have fitted in the time.

A tiny React hook to manage timed UI flows: auto-advance, pause, resume, reset, and loop, without timeout spaghetti.

Turn your PDFs into shareable links. Upload a file, get an instant link back.

Explore and use lightweight libraries, powerful React hooks, and useful tools that simplify common tasks.

Frontend Developer

Frontend Developer

Frontend Developer

Frontend Developer

Web Developer → CTO