Factory AI is building 'Droids' — autonomous software development agents that automate various engineering tasks like PR reviews, feature implementation, and incident fixing for enterprise teams. They are model-agnostic and integrate across multiple surfaces like CLI, IDE, and CI/CD. The company has significant funding and a high valuation, serving major enterprise customers.
A starter prompt for Claude Code, what you'll need, and how to reach them.
You are an expert AI engineer and a solo developer. Your goal is to build a minimal viable product (MVP) of an autonomous software development agent, specifically focused on automating code refactoring for a single, well-defined pattern. This MVP will be a proof-of-concept demonstrating the core 'Droid' capability of Factory AI. Build this using Next.js 16 (App Router), React 19, Tailwind v4, and AI SDK v6 with Gemini. The backend will be Node.js and a Neon Postgres database for task persistence.
Here's the plan:
1. **Agent Core:** Develop a Node.js backend service that exposes an API endpoint `/api/refactor`. This endpoint will accept a GitHub repository URL and a specific refactoring pattern (e.g., 'replace all instances of oldFunction() with newFunction(arg1, arg2)').
2. **Refactoring Logic:** Implement the agent logic to:
a. Clone the specified GitHub repository locally (or via GitHub API).
b. Use a Gemini Pro LLM via AI SDK v6 to identify all instances of the 'oldFunction()' and generate the necessary code changes to replace them with 'newFunction(arg1, arg2)'. Ensure the LLM output is structured for programmatic parsing (e.g., diff format or JSON describing changes).
c. Apply these changes to the cloned repository.
d. (MVP stretch goal) Generate a new Git branch and open a pull request with the refactored code.
3. **Frontend (Next.js):** Create a simple Next.js web application with:
a. An input form for the GitHub repository URL and the refactoring pattern.
b. A button to trigger the refactoring process.
c. A display area to show the status of the refactoring and, if successful, a link to the generated pull request (or a diff).
4. **Database (Neon Postgres):** Store a record of each refactoring task, including input parameters, status ('pending', 'in progress', 'completed', 'failed'), and output (e.g., PR URL, error message).
**Build Gate:** A user can input a public GitHub repository URL and a simple refactoring pattern, and the system successfully generates and displays a diff (or opens a PR) showing the proposed changes. Prioritize robust error handling and clear status reporting for the MVP. Assume API keys are environment variables.The audience interested in Factory AI's autonomous dev agents are developers and engineering leaders who would value tools like `forge-kit` for automating their build loops, `repo-gardener` for maintaining code health, and `ai-usage-monitor` for managing AI tool spend.
Factory AI | AI, Fullstack, Frontend, Platform, Data, Product Engineers + GTM, Design, Operations | San Francisco, CA (+ NY, London for some roles) | ONSITE | Full-time | https://factory.ai At Factory we build Droids, autonomous software development agents that review PRs, implement features, run tests, fix incidents, and handle migrations for enterprise engineering teams. Droid is model agnostic and surface agnostic, and works across your CLI, IDE, Web/Desktop app, CI/CD, Slack, and Linear. Customers include Nvidia, Adobe, Adyen, Klarna, and Ernst & Young. We raised a $150M Series C at a $1.5B valuation, backed by Khosla Ventures, Sequoia, Blackstone, Insight Partners, NEA, and NVIDIA. We're hiring across the stack: - AI Engineer: agentic systems, LLM integrations, evals - Fullstack Engineer: platform, APIs, auth, billing, terminal UIs - Frontend Engineer: web and desktop apps (React, TypeScript, Electron) - Platform Engineer: infrastructure, sandboxing, CI/CD, cloud - Product Engineer: end-to-end feature ownership - Data Engineer: pipelines, analytics, telemetry at scale - Security Engineer / Research Engineer Stack: TypeScript, React, Node.js/Bun, Python, GCP/AWS, Firestore, E
Required for GitHub API interactions (cloning, opening PRs).
Standard AI SDK v6 integration.
Standard database for Next.js apps on Vercel.
Deployment platform for Next.js application.
Apply via the careers page on https://factory.ai/careers or use the contact information provided on their 'Who is Hiring' post if available (not in this snippet).
“I'm a solo operator deeply interested in autonomous developer agents. I've built an MVP prototype demonstrating an AI agent capable of automated code refactoring. I'd love to share how I approached the problem and discuss how my skills in AI engineering and full-stack development align with Factory AI's mission.”
Open the original ↗