Forward Deployed AI Engineer

Avenue Z•Miami, FL

8h•$90,000 - $100,000•Onsite

About The Position

We are hiring a Forward Deployed AI Engineer to build and ship the agent systems that run social commerce for our brands and for our own platform. You will embed with the teams doing the work, performance media, creator and influencer, ecommerce, and analytics, find the process where throughput is lost or money is left on the table, and build a production agent that owns it. Then you move to the next one. Concretely, this is work like: agents that detect and re-authorize expiring TikTok Shop affiliate codes and Meta partnership permissions before spend stalls, systems that reallocate ad budget toward top-performing creators automatically, pipelines that launch new creator content across every platform daily without a human touching it, and reporting infrastructure that tells a brand what actually drove GMV. You own the problem end to end: discovery, architecture, build, deploy, evals, adoption, and what breaks at 2am. If your instinct on a messy problem is to open an editor rather than schedule a working session, this is your role.

Requirements

Computer science or another technical background, or equivalent hands-on experience as a developer or as a self-taught AI-native builder. We care about what you have shipped and whether it works, not where you learned it.
3 to 5 years building and shipping software that other people depended on.
You can read, write, and debug Python or TypeScript. You have deployed and maintained real services and you understand what is happening underneath, whether you wrote every line yourself or built it alongside AI coding tools.
Hands-on experience building LLM agent systems: function and tool calling, RAG, orchestration, prompt and context engineering, agent evals.
Real experience integrating third-party marketing, ad, or commerce APIs. You know what OAuth token expiry, rate limiting, webhook replay, and undocumented API behavior actually cost you.
Comfortable with cloud infrastructure (AWS or GCP), containers, CI/CD, and SQL.
You work fluently with AI coding tools such as Claude Code and Cursor. We expect you to build with them. We also expect you to be able to explain, debug, and defend what comes out of them.
You understand how performance marketing works: ROAS, CAC, attribution, creative testing, funnel mechanics, affiliate and creator economics. Or you can learn a domain fast and go deep, and you can prove it.
Entrepreneurial. You have built something of your own, a startup, a product, a side venture, an open source project, and can point to what you shipped and what it did.
High agency. You gather your own requirements, make the call, and ship. You do not wait for a spec, a committee, or a change management plan.
Able to sit with a senior executive or a brand's head of ecommerce, understand the business problem in their language, and translate it into a system without a translator in the room.

Nice To Haves

Direct experience with TikTok Shop, Meta Ads, or Amazon Seller/Ads APIs.
Background at a DTC brand, an ecommerce platform, a creator or affiliate network, or a performance agency.
Prior forward deployed, applied AI, or founding engineer experience.

Responsibilities

Embed with a business unit, diagnose where throughput or margin is lost, and ship a working agent against it inside the first few weeks.
Build production agent systems: tool use, planning loops, memory, retrieval, orchestration, and human-in-the-loop approval gates where money or brand safety is on the line.
Integrate deeply against the commerce and ad platforms our revenue runs on: TikTok Shop and TikTok Ads, Meta and Instagram, Amazon, Shopify, Google, Snap, and affiliate and creator APIs.
Build data pipelines and analytics infrastructure that tie creator content, ad spend, and attributed GMV into one picture.
Build the evaluation harness alongside the agent. Nothing ships to production without a way to measure whether it is right.
Own the infrastructure: APIs, queues, auth and token refresh, observability, error handling, rate limits, and cost controls.
Instrument everything. Every system you ship gets tied to a number: hours returned, cycle time cut, spend efficiency, GMV or revenue supported.
Document it, drive adoption with the team that uses it, and keep improving it after launch.