Software Engineering Intern (AI/ML). We’re building production AI agents that automate outbound sales end-to-end. If you’re early in your career and already thinking in: Python and LLM APIs Prompt engineering Agent pipelines and async workflows Breaking systems before customers do If you want to build real AI infrastructure early in your career, let’s talk. Details in the attached JD.
Roister is replacing B2B sales teams with autonomous AI agents – systems that find prospects, write emails, handle replies, and book meetings without a human in the loop. We’re not demoing this. We have paying customers.
You’ll sit next to our full-time engineer and work on the real product. The code you write will run in production. The prompts you test will go out to real campaigns. The bugs you catch will matter.
If you want to understand how AI agent systems actually get built, this is where you learn that.
What you’ll actually do:
- Keep the Python agent backend running and make it better
- Write and test prompts that power live outbound sales workflows
- Break the agent pipelines before customers do
- Build the tooling that makes our engineer faster
- Span the full stack when needed — Next.js on the front, Python on the back
- Watch agent performance in Langfuse and speak up when something’s off
What you need:
Python you’ve used to build something real. Experience hitting LLM APIs — OpenAI, Anthropic, Gemini, doesn’t matter which. Enough REST API fluency to read the docs, make the call, and debug when it breaks. The ability to work without someone checking on you every hour.
What helps:
You’ve touched LangChain, LangGraph, or any agent framework. You know what a system prompt actually does. You can read a JSON payload and tell what’s wrong with it. React or Next.js doesn’t scare you.
What matters most:
You finish things. You’d rather ship something imperfect than polish something nobody’s waiting for. When something breaks, you get curious, not stuck. You take the work seriously even when nobody’s watching.
Stipend: Yes
Duration: 3–6 months
Location: Remote (with real overlap in EST/PST for team syncs)
