AI Product Engineer, Full-Stack (Applied AI / Agents)

Compensation: $50k – $85k • No equity

About the role Wired2Scale manages billions in client ad spend. Behind that spend is a tech problem most companies never face: automating decisions on accounts spending $1M+ a day, building the dashboards that make sense of it, and giving media buyers tools sharp enough to act in real time.

You'll build W2S SI, the systems our buyers run on. A buyer is up at 1 AM deciding whether to kill an ad or scale it, and the tool you built is what they're looking at. When your code breaks, someone's account breaks with it. We're hiring one engineer who wants that weight.

What you'll work on

• Automated media-buying rules: killing, scaling, pausing and moving ads and budgets on Meta (plus Google Ads, AppLovin, TikTok and Snap), with rules customized per client.
• Data pipelines and unified dashboards: combined reporting from Meta, Google, Northbeam, Triple Whale and Shopify. Data has to be complete and certified: partial pulls never reach a buyer.
• Agentic tooling: a chat interface (Slack/desktop) that lets buyers and strategists take real actions in one place, plus MCP/API integrations so AI agents can safely operate ad accounts.
• Strategy systems: KPI target dashboards, scale threads, account-structure tools and forecasting, the intelligence layer our strategists use to make calls in real time.
• Forward deployment: take systems that work on one account and make them work on every account. Sit with strategists, find what breaks, and close the gap.
• Safety by design: read-only defaults, approval gates, audit logs and fail-safe behavior. Every automated action has to answer "what happens if this data is wrong?"

What we're looking for

• 3+ years shipping production software, strong in at least one of Python or TypeScript (Go, Rust or Java are fine if you know your tools deeply).
• You understand how LLMs, embeddings, vector databases and AI agents work under the hood, not just well enough to call the library.
• You've shipped AI software people actually use: in production, open source or serious side projects.
• Solid backend and data fundamentals: relational schema design (Postgres), idempotent and observable data pipelines, and API integrations with rate limits and partial failures.
• You drive AI coding tools (Claude Code, Codex, Cursor) and you review, test and verify what they produce. Not a "vibecoder".
• You write and speak clear English. Most work is async, so your writing has to carry what a meeting would.
• Self-directed and resourceful: missing context, you go get it; blocked, you find a workaround, tell the team, and help fix it properly. (If you need a manager to unblock you every morning, this will be frustrating.)

Nice to have

• Experience with the Meta Marketing API, Google Ads API, TikTok, Snap or AppLovin ads APIs, or with attribution tools (Northbeam, Triple Whale).
• Performance-marketing or e-commerce/DTC domain knowledge (CPA, ROAS/MER, CAC, budgets, creatives).
• Supabase, Next.js/React, Vercel, Linux VPS/DigitalOcean, Docker.
• Building MCP servers, OAuth integrations or multi-agent systems.
• Front-end/UI sense for internal dashboards.

Our stack Supabase (Postgres, auth), Next.js/React on Vercel, Python and TypeScript services, DigitalOcean droplets, Redis. Integrations: Meta, Google Ads, AppLovin, TikTok, Snap, Northbeam, Shopify, Slack. AI tooling: Claude / Claude Code, Codex, Cursor, Grok, MCP servers.

How we work

• Fast. Ship a v1, get feedback, iterate. Code still has to be maintainable and has to tell people what it's doing.
• You actively contribute and prove your value. You start where your skills are and move up based on what you ship.
• Remote-native and async-first, with a team across the US, Europe and Asia.

Hiring process

1. Apply with your answers to the screening questions, plus links to GitHub or shipped work. A short (2-3 min) video intro is strongly encouraged.
2. Intro call with Cody (W2S SI) and Shiv (Lead Engineer): background, how you use AI tools, culture fit.
3. Take-home assignment (short, practical; e.g. design a Meta + attribution data pull and storage layer), followed by a technical deep-dive walking through your trade-offs and failure handling.
4. Final interview with Tim (leadership).
5. Offer, then a one-week trial working on a real task with the team.