# Software Engineering Intern - Unbound
## The Problem
AI coding agents (Claude Code, Cursor, Copilot) can write code, run shell commands, and modify infrastructure across an entire engineering org. Most companies have zero visibility into what these agents are doing and no way to stop a bad tool call before it executes.
We're building the governance and security layer for AI coding agents: agent and tool discovery across dev fleets, usage and token analytics through an LLM gateway, and runtime policy enforcement that intercepts risky actions before they happen.
## Who We Are
We're [Vignesh](https://www.linkedin.com/in/vigneshsubbiah16/) and [Raj](https://www.linkedin.com/in/rajaramsrinivasan/). We come from deep engineering, product, and security backgrounds. We started Unbound because we kept seeing orgs hand AI agents broad access to codebases, infrastructure, and APIs with zero guardrails. That felt like a problem worth solving. We're YC-backed and building out of the Bay Area.
## What You'll Do
**Ship real features.** You'll own features from design through deployment. Your code runs in production and gets used by engineering teams at real companies.
**Work across the stack.** Our product spans agent discovery, an LLM gateway, and a runtime policy engine. You'll touch Python, Next.js, Kubernetes, and cloud infra. If you haven't used something on this list, you'll learn it here.
**Work with founders directly.** You'll have full context on what we're building and why. You'll sit in on customer calls, see the roadmap, and understand the business, not just your tickets.
**Pick up what matters.** No backlog grooming ceremonies. You'll see gaps, flag them, and fill them.
**Build in AI security.** You'll work hands-on with how AI agents operate under the hood (hooks, MCP servers, tool-use lifecycles) and build the guardrails around them. This space barely existed a year ago.
## Your First 30 Days
In your first two weeks, you'll ship a feature to production. Past projects have included agent integrations, policy engine rules, and dashboard features. Your project depends on what's highest priority when you start, but it will be real and it will ship. By day 30, you'll have full context on the product, customers, and codebase, and you'll be picking up work on your own.
## What We're Looking For
* **You can ship.** You've built and deployed something: a side project, a class project, a startup, open source. The bar isn't perfection, it's follow-through.
* **You're comfortable with ambiguity.** We'll give you a problem, not a spec. You figure out the approach, ask good questions, and iterate.
* **You code with AI tools and stay in control.** We expect you to use Copilot, Claude, whatever works. But you understand what the code does and can debug it when it breaks.
* **You're curious about security, devtools, or infrastructure.** Deep experience not required, but genuine interest matters. If you've thought about what happens when an AI agent has root access to a codebase, we should talk.
* **You want to understand how the whole system works,** not just your piece of it.
## Nice to Have
* Experience with Python, TypeScript/Next.js
* Familiarity with cloud infrastructure (Azure, AWS, K8s)
* Experience building in the AI/LLM stack
* You've been a founder or want to be one
## Our Stack
Python · Next.js · Azure · Kubernetes · OpenAI · Anthropic · Google Vertex AI
## How We Work
* **Small team, high trust.** No micromanagement. We trust you to manage your time and deliver.
* **Async-first.** We don't burn your day with meetings. Communication happens in writing, calls happen when they need to.
* **Direct feedback.** We tell you what's working and what isn't. We expect the same back.
* **Ship > talk about shipping.** We'd rather see a rough working prototype than a polished slide deck.
Visa sponsorship is not available for this role.
