About Superpower The comment section under a paid ad changes how that ad performs. Superpower runs those comment sections for brands. Our AI moderation hides what hurts the ad. Our reply engine answers customers in the brand's voice. Our DM agents pick up conversations that start in the comments, and a workflow builder connects it all to other tools.

We work with brands like WestJet, Baskin Robbins, Skool, Scribd and Acquisition.com, and we're three people. Next, we're building infrastructure for brands with large UGC creator programs.

The role

You'll be our first engineering hire, working directly with Justin (cofounder, technical) and Alan (cofounder, everything else). There's no team to hand things off to. You'll take problems from "a client is seeing this" to shipped and running in production, and you'll decide much of the how yourself.

What you'll work on

- AI systems that make decisions in public: whether to hide a comment, what to reply, when to hand off to a human. Wrong answers show up under a brand's ad.
- Keeping our pipelines reliable against Meta's APIs, including rate limits, spam detection, and account connections that drop.
- Evaluation: proving a prompt or model change made things better before it reaches every client.
- The app itself, which our team and our clients' teams work in every day.

What you need coming in

- Strong TypeScript, backend and frontend. You've shipped Node.js APIs and React apps that real users depended on.
- Solid SQL and relational schema design, including migrations against a live production database.
- You've built something with LLMs that real people used, and dealt with what happened when the model got it wrong.
- You've integrated third-party APIs and handled their failures: flaky endpoints,



missing data, bad documentation.
- You test your work and care whether it holds up after it ships.

What you'll own and grow into (we don't expect all of it on day one)

- Background jobs and queues done properly: retries, idempotency, failure handling
- Prompt engineering as engineering: versioned prompts, structured outputs, tool calling, handling bad model responses
- Retrieval and knowledge design: how an agent finds the right information at the right time, choosing the approach that fits the problem instead of defaulting to one
- Deciding when the system acts on its own, when it asks a human, and when it stays out of the way
- Evaluating model changes with evidence: test sets, regression checks, measurable quality
- Designing against the ways LLM systems fail in customer-facing products

Nice to have

- Experience with Meta, Instagram or TikTok APIs
- Experience building visual editor UIs (our workflow builder is node-based)
- Background in customer support, social media or marketing tooling
- Design sense. We don't have a designer.

What it's actually like This is a small team doing a lot, and the hours are long. Scope changes when customers teach us something, sometimes week to week. You'll talk to customers, fix things that aren't technically "your" problem, and make calls without anyone to check with first.



If you want a well-defined role with transparent boundaries, this isn't it, and that's fine.

In return, you get real ownership of a production system with paying customers, equity in the company, a voice in every decision, and the fastest way there is to learn how a software company gets built.

Pay

Salary plus equity. The salary is below what larger companies pay, and the equity is meant to make up the difference. We share the full numbers, salary and equity, in the first conversation, so nobody wastes time.

Experience

There's no required number of years. We care about what you've built. If you're early in your career, show us evidence: something you shipped, people who used it, a hard problem you solved.

How to apply

Apply via this post, and then send three answers to the questions below to [email protected]:

1. A link to something you built and shipped (a product, repo, demo or write-up), and which part was yours.
2. A paragraph or two answering this: Our AI writes replies to public comments on brands' paid ads. What's the first thing you'd worry about going wrong, and how would you catch it before a client does?
3. A paragraph or two answering this: Tell us about something you started or took over with nobody asking you to, like a product, a side business, a club, or a broken process at work. What did you do, how did it turn out, and what would you do differently?

Please refrain from writing this entirely with AI, we read every single one of these, and we'll know right away. We would rather you send something shorter that you wrote yourself, vs. longer and AI written. We won't review applications that are missing an email.

📌 Founding Engineer, Full Stack + AI (Toronto)
🏢 Superpower
📍 Toronto