About RiverX
RiverX is an AI platform for building and evolving websites and web applications. It brings model selection, coding agents, repository-backed projects, live previews, database provisioning, and publishing into a connected workflow. People should be able to move from an idea to a working application, understand what the agent changed, and keep improving it without losing control of their code.
Our product spans a Next.js and React frontend, a Node.js backend, agent and workspace services, Supabase for authentication and core application data, and Turso for project databases and the skills catalog. GitHub, Railway, Vercel, and external model providers are part of the engineering environment. Each role below owns a specific part of making that experience dependable.
The opportunity
Explore how coding agents behave on real application-building tasks. You will help make RiverX’s AI workflows more dependable by preparing useful examples, examining failures, and testing focused improvements. This internship is suitable for someone who enjoys both writing software and asking whether an AI system actually completed the task it was given.
What you will work on
- Create small, reproducible evaluation tasks based on workflows such as editing a React page, fixing an API bug, or connecting a form to data.
- Define what success means before running an experiment: correct behavior, a passing check, a usable interface, or a clearly stated limit.
- Review agent traces and code changes to identify missed context, incorrect assumptions, tool errors, unnecessary edits, and weak validation.
- Compare prompts, context selection, tool descriptions, or model behavior on the same set of examples.
- Build simple scripts to organize results and make experiments repeatable, keeping failures and regressions visible alongside improvements.
- Learn how reusable skills, repository context, conversation history, and tool results influence a coding run.
- Work with engineers on bounded changes and verify that an improvement holds beyond the single example that motivated it.
- Keep repository content, model output, and external instructions separate from the user’s actual intent and permissions.
What you should bring
- Ability to write and debug software in JavaScript/TypeScript or Python, and willingness to read a Node.js application codebase.
- Some experience using a model API or building an AI-assisted application, experiment, or tool.
- Basic familiarity with Git, running an application locally, and checking whether a code change works.
- Curiosity about failure analysis and the discipline to record an unexpected outcome instead of silently discarding it.
- Clear writing: explain what you tried, what you observed, and which conclusions the evidence supports.
- Comfort asking questions about assumptions and working through a problem whose answer is not known in advance.
What you can learn
Evaluation design, coding-agent tool use, context budgeting, prompt iteration, trace analysis, and the difference between a convincing response and a correct implementation. A research degree or prior model-training experience is not required.
A useful first project
Prepare a small task set with repeatable inputs and explicit acceptance criteria. Identify one recurring failure pattern, test a focused change, and write a short report that includes both improvements and cases that remain unresolved. Prefer a clear result on a modest task set to unsupported claims about general agent intelligence.
Include in your application
An AI experiment, software project, or evaluation you created. Describe how you judged its output, one thing that failed, and what you would test next. Include runnable code or a concise walkthrough where you can share it.
Internship arrangements
This is a remote internship. Please share your location, time zone, earliest start date, weekly availability, and any academic commitments in your application. Duration and working arrangements will be discussed during the application process rather than assumed from a standard academic calendar. The role is intended for students, recent graduates, and early-career applicants; prior full-time employment is not required.
You will work on scoped tasks, make your assumptions explicit, and ask for review before a change reaches users. We value clear progress updates, willingness to use feedback, and careful verification more than familiarity with every tool in the stack. This listing does not promise conversion to a full-time position.
How to apply
Apply through this page with your résumé as a PDF and a short note about why this work interests you. Include relevant project, writing, portfolio, or campaign links and explain what you personally contributed. Coursework, community work, independent projects, and open-source contributions are welcome evidence. Share only work you have permission to share; do not include confidential customer information, credentials, or private business records.