Position overview

Matrix is building its product capability in house around small, high-velocity teams using AI-first workflows. We are seeking a high-aptitude university graduate to join as a Junior Software Engineer. Unburdened by traditional software engineering patterns, you will leverage modern AI tools to build and ship real software from day one.

Working alongside senior engineers and product builders, you will develop Next.js, TypeScript, and Node features deployed on Vercel and Postgres. Using tools like Claude Code and Cursor as your core environment, your primary role is directing AI agents, critically reviewing generated code, and ensuring product alignment rather than writing manual syntax.

We prioritize attitude, curious reasoning, and a product-focused mindset over formal technical experience. You will participate actively in product discussions, storymapping, and user experience decisions, focusing on who the software is for and why it matters.

This is a small team, so there is no large engineering group to hide inside, and the support you get will be direct and frequent rather than formal. 

Key responsibilities

1. Build and ship features in a pair

• Build Next.js and Node features, from components and API routes through to a deployed preview, using Claude Code and Cursor as your primary tools
• Work through the full change with a senior engineer: write it, test it, get it reviewed, get it merged, watch it in production
• Write tests for what you ship, whether you wrote the coverage yourself or generated and checked it
• Read existing code and system architecture prior to making changes to maintain seamless integrations across core platforms like Salesforce and Snowflake

2. Learn to review agent output properly

• Treat all agent-generated code as a first draft: systematically review line-by-line, detect edge cases, identify logic or security gaps, and refine the output
• Master advanced prompting strategies, agent steering, and context management to continuously optimize the quality and precision of AI output
• Ask when something an agent produced works but you cannot explain why it works, because that gap is the thing to close
• Build your own picture of where these tools are reliable and where they are not, and share it with the team

3. Take one thing end to end, including the product side

• Own a small internal tool or feature from brief through to rollout and the first round of changes based on real usage
• Sit in on discovery and storymap sessions for the products your team is building, and contribute the feasibility view as you learn it
• Check how what you shipped is actually being used and report back what the numbers show

What we’re looking for

Experience and qualifications

• Recent university graduate in Computer Science, Software Engineering, or a related field with strong foundational problem-solving abilities
• Demonstrated exposure to modern web technologies (JavaScript, TypeScript, React, Next.js, or SQL) through university coursework, personal projects, or self-directed learning
• You can read and write SQL well enough to work with a relational database
• Adaptable, open mindset free of legacy software engineering habits, ready to adopt AI-native development paradigms from day one
• You can explain your own code to someone who did not write it

AI fluency and builder mindset

This matters to us. You don’t need professional experience with agentic tools, but you do need to be using them already and thinking critically about what comes back.

• You use Claude Code, Cursor or a comparable tool regularly and have started forming opinions about what makes a prompt work
• You have caught an agent being confidently wrong and can describe what tipped you off
• You build things because you want to see if they work, not only because someone set the task
• You are curious about how software actually reaches production, not only about the code itself

Skills and attributes

• Interested in who the software is for. You want to know what a student or a parent does with it
• You ask early rather than staying stuck, and you come with what you have already tried
• You take feedback on your code as information rather than criticism
• Clear written communication, because most of the context you give an agent is written and the same clarity matters in a pull request
• Honest about what you don’t know yet, which at this stage is the most useful trait on this list

Nice-to-have

• Any exposure to CI/CD, testing frameworks or deployment tooling
• A public repository, a project write-up, or anything that shows how you think
• Interest in education, or experience tutoring or teaching

Success measures

Success in the first 90 days

• You have shipped multiple changes to production, reviewed and merged, with no unresolved issues left behind you
• You can take a well-defined ticket from start to deployed without step-by-step direction
• You can show a specific case where you rejected or rewrote agent output and explain why
• You have attended discovery or storymap sessions and asked useful questions in them

Success in 6 to 12 months

• You have owned one internal tool or feature from brief through to rollout and a second round of changes based on how it was used
• You work through larger changes independently, bringing questions to a senior engineer rather than needing the work broken down for you
• Your reviews of other people’s work, and of agent output, are useful to the team
• You can explain how what you built affected a number the team cares about