👾 About the Job

Company: AdsGency AI

📍 Location: Onsite / SF Bay Area Preferred

💼 Employment Type: Full-Time

🚀 About AdsGency AI

We’re AdsGency AI — a hacker-founded startup building autonomous, multi-agent systems to run digital advertising with zero human input.

Our product is a real-world application of AI agents that write ad copy, generate creatives, optimize budgets, launch campaigns, and learn from performance data — across Google, Meta, TikTok, and beyond. Think: LLM agents that ship ROAS.

We move like hackers, build in public, and write code that automates the world.

🧠 The Role – Member of Technical Staff, AI Agent (Multi-Agent Systems)

You’ll be one of our core engineers, helping us build a deeply technical and highly autonomous agentic platform. You’ll work across the full stack — from LLM orchestration and decision logic to real-world action-taking agents in production.

This is not a model research role. It’s not just prompt tuning. It’s real-world agent engineering:

LLMs with tools, memory, agency, and impact.

🔧 What You’ll Build

• 🤖 Agentic Ad Infrastructure – Build LLM agents that autonomously launch, optimize, and iterate ad campaigns

• 🧩 Multi-Agent Architectures – Use (or replace) frameworks like LangGraph, CrewAI, AutoGen, etc.

• 🧠 LLM-Tool Interfaces – Create agents that invoke APIs, run retrieval, self-reflect, and write to real systems

• 🛠 Agent Memory & Planning Systems – Integrate short-term and long-term memory, tools, tasks, and loops

• 🧪 Debugging & Evaluation Tools – Build in public, with trace-based evaluation for chain-of-thought reasoning

• 💥 Real-World Impact – These agents don’t just reason. They ship ads, generate revenue, and outperform humans.

💡 What We’re Looking For

✅ You’ve built with LangChain, CrewAI, AutoGPT, or your own agent framework

✅ You’ve used LLMs as coworkers (Cursor, Copilot, ChatGPT, Claude, Tabnine, etc.)

✅ You’ve shipped something weird, useful, or deeply technical with LLMs

✅ You’re not afraid to scrap the framework and write your own orchestration engine

✅ You’ve deployed agents in production, not just in Colab

✅ You like CLI tools, traces, logs, and think about agents as systems with state

🧠 Bonus Hacker Energy

• You’ve scraped something and built a data pipeline in a weekend

• You write shell scripts to automate your daily workflow

• You use git blame to learn and ship faster

• You’ve made PRs to open-source LLM tools or reverse-engineered APIs

• You believe memory, evaluation, and agent coordination are the real frontier

⚙️ Tech Stack We Love

• Python, FastAPI, LangChain, LangGraph, TypeScript, React

• OpenAI, Claude, HuggingFace, Ollama

• Vector DBs (Weaviate, Qdrant, LanceDB), Redis, Postgres

• Docker, AWS, Supabase, Railway

• GitHub > Slide decks

• Logs > Dashboards > Charts > Fancy reports

🎯 Who You Are

• Hacker > Engineer > Researcher

• 0-to-1 builder who ships scrappy before elegant

• Comfortable in open-ended, ambiguous environments

• Writes code in public, learns fast, fails fast, and debugs hard

• Interested in real-world agent deployment, not just theoretical chains

💰 What You Get

• Competitive salary + real equity

• Founding-level technical ownership

• Direct access to revenue-generating use cases (ads that work)

• Hacker-first culture with full engineering autonomy

• Early-stage engineering imprint on an AI-native company

• A team that writes code, not policies