Senior Full-Stack Engineer (AI-Native)

I'm Connor, founder of AdStream. I'm based in Australia and I run this company small and fast.

AdStream manages the ad creative workflow for e-commerce brands running creator-made (UGC) video ads at volume: creative briefs go out, creators upload video, the brand's team reviews and approves everything in the browser, and approved ads publish straight to Meta.

See it for yourself before you apply. Make an account at ---------- , click around, poke at the brief and review flows. The best applications come from people who've actually touched the product.

What we're building right now
Video playback. Reviewers play and scrub large creator videos in the browser, and right now playback is laggy. We serve full-quality source files straight from S3, so reviewers wait on buffering instead of reviewing.

The route forward is Drive-first: Google Drive becomes the home for every customer asset. That gets us three things at once, fast playback on large files, customer assets living in their own Google Drive (a place they control and a brand they already trust), and the foundation for semantic search. That last one is the exciting part: once every file lands in Drive, we process it, get the transcript and text embeddings so you can search what's said in a video, frame embeddings so you can search what's visually in it. "Find b-roll with a kettle on a table." "Pull every clip where someone says sustainable." Internal teams find images, b-roll and past ads in seconds instead of scrubbing through folders, which makes video editing and graphic design dramatically faster.

Beyond that, the roadmap you'd own: deepening the Meta integration and adding TikTok (OAuth, campaign publishing, status webhooks), AI creative generation with OpenAI, and Stripe billing.

The stack
Next.js 15 (App Router, server actions) + React 19 + TypeScript. Tailwind v4. Supabase — Postgres with row-level security, auth, storage. AWS S3. Google Drive API. Meta Graph API. OpenAI. Deployed on Vercel.

You don't need prior experience with every line, but be honest about which parts you haven't touched, and don't propose a rewrite in week two.

This is an AI-native role
We provide a paid Claude Code seat and cover your AI tooling. You must already work this way daily, agentic tools (Claude Code, Cursor or equivalent), not autocomplete. You can write a spec an agent can execute, review agent output critically instead of trusting it, and verify AI-written code before it ships. One senior with a real agentic workflow now ships what a small team shipped a few years ago, that's the bet this role is built on.

How we work:

A 10–20 minute call every morning: what's planned today, review of yesterday, check in on deadlines.
We agree on milestones upfront so everyone stays accountable.
Be upfront about how long things will take. Not so we can micromanage, so we can be honest with our users and set expectations they can trust. Don't tell us it's a day if it's a week.
Own outcomes, not tasks. Questions early, blockers immediately.
Async outside the morning call: Looms, written updates. Hours are 9am–5pm AEST — a normal 7am–3pm day shift in the Philippines, evenings free.

How to apply:

Start your application with the exact phrase "Drive-first." It tells me you actually read this.

Record a Loom (5 minutes max — clear and tight beats thorough and rambling) covering:
You and your work. Quick intro, then walk me through something real you've shipped, what it does and what you built.
AI doing meaningful work. One real example of using an agent for something substantial: how you directed it, something it got wrong that you caught, and how you verified it before shipping. I want to see judgment, not AI slop.
How AI changed your workflow. How much more do you ship now? Do you still write most lines by hand, or mostly review and direct?
Your models. Which model do you prefer and why? Do you run more than one, and how do you split work between them?

If your Loom lands, we'll jump on a call.

Looking forward to hearing from you!