Principal Engineer - AI AdTech Media Buying Platform

Three Pillars Recruiting; United States; Vor Ort; USD 180.000 - 280.000; Vollzeit

Zusammenfassung

Three Pillars Recruiting seeks a Principal Engineer to own end-to-end areas of the product, spanning the agent harness, evaluation systems, platform integrations, and the user interface. You will work on a lean team with engineering autonomy, from idea to production, and directly with customers as media buyers use the systems you build.

You will integrate a new advertising platform, design robust APIs, and drive ownership from design through production, with strong emphasis on reliability and

Qualifikationen

• 5+ years backend engineering experience.
• Production Python experience or equivalent depth in another language.
• Comfort with asynchronous systems, Postgres, schema design and background jobs.
• Hands-on experience with LLM agent systems and tool calling.
• Diagnose nondeterministic systems using evaluations, replay, tracing, and telemetry.
• Frontend experience with TypeScript and React; Next.js and Tailwind helpful.

Aufgaben

• Integrate a new advertising platform.
• Own an agent-powered product surface with distinct subsystems.
• Eliminate recurring production failures using logs and traces.
• Extend the agent harness with improved tooling and review systems.

Kenntnisse

Backend engineering

Python

Async systems

Postgres

Schema design

Background jobs

Distributed systems

LLM agent systems

Tool calling

Context management

Structured outputs

TypeScript

React

Tools

Google Cloud

Vertex AI

Cloud Run

BigQuery

Pub/Sub

FastAPI

pgvector

Redis

Next.js

Tailwind

CopilotKit

AG-UI

Jobbeschreibung

Principal Engineer

You will own meaningful, end-to-end areas of the product, spanning the agent harness, evaluation systems, platform integrations, and user interface. They are a lean team with minimal hierarchy, significant engineering autonomy, and a fast path from idea to production.

The development workflow is agent-native: coding agents are embedded throughout exploration, implementation, testing, review, and debugging, while engineers remain accountable for the quality of what reaches production. You will also work directly with customers and observe media buyers using the systems you build.

What You'll Do

Integrate a new advertising platform.

• Design and implement authentication, a robust API client, read and write tools with appropriate approval controls, a synchronization pipeline into Postgres, and a dedicated platform sub-agent. You will also develop model instructions that capture the platform's real-world API behavior, including nuances and constraints that may not be fully reflected in its documentation.
• Own an agent-powered product surface. Media-plan imports, campaign building and quality assurance, scheduled briefings, and Slack and email entry points are each distinct products with their own agent architecture, tools, evaluations, and failure modes. You may take ownership of one from design through production.
• Eliminate a recurring class of failure. When something is silently going wrong in production for a particular customer, you will reconstruct the agent's behavior using BigQuery event logs and traces, identify the underlying mechanism, and address the root cause.
• Extend the agent harness itself. Improve foundational capabilities such as tool gating, context compaction, large-payload handling, and the review system that determines whether the operator's work is ready to present to a user.

What They're Looking For

• 5+ years backend engineering experience. You have several years of production Python experience, or equivalent depth in another language that you can transfer quickly. You are comfortable with asynchronous systems, Postgres, schema design, background jobs, and the practical failure modes of distributed systems.
• Hands-on experience with LLM agent systems. You have built and shipped an agentic product involving tool calling, multi-step orchestration, context management, and structured outputs. You also treat prompts and instructions as versioned, testable engineering artifacts.
• Comfort with nondeterministic systems. You can diagnose systems in which the same input may produce different outputs. You use evaluations, replay, tracing, and telemetry to understand behavior.
• Enough frontend experience to complete the feature. You are comfortable with TypeScript and React and can ship the interface required for a feature without handing it off. Prior experience with Next.js and Tailwind is helpful but can be learned on the job.
• A strong sense of ownership. They are a small team responsible for a large product surface. You take responsibility for carrying work from an initial idea through implementation, validation, and production.

Preferred

• Experience in advertising technology or marketing technology, including DSPs, ad servers, attribution, or the structure of media plans. This domain knowledge is difficult to find, and they value it highly.
• Deep experience with Google Cloud, particularly Vertex AI, Cloud Run, BigQuery, and Pub/Sub.
• Experience building multi-tenant SaaS products with rigorous data-isolation requirements.

Their Stack

Python

• Google ADK
• FastAPI
• Postgres + pgvector
• Redis
• Vertex AI
• BigQuery
• Cloud Run
• Pub/Sub
• Tofu
• Next.js
• React
• TypeScript
• Tailwind
• CopilotKit / AG-UI