As an AI Engineer, Software Development Life Cycle, you build and improve the systems that allow AI to generate and evolve software.

Your goal is to make AI-driven development reliable, fast, and scalable.

Responsibilities

Your work includes:

• building and improving our context and harness for coding agents
• developing prompts, skills, and automation workflows
• designing development pipelines and tooling
• creating evaluation frameworks
• testing new models and capabilities

You continuously experiment and iterate to improve automation performance.

We want to use the best coding agents on the market and make them build the Pit way.

What Success Looks Like

• coding agents reliably generate high-quality systems
• automation significantly reduces development time
• improvements in models translate into real productivity gains

Qualifications

Required

• 5+ years of professional software engineering experience spent building production systems in TypeScript/Node.js
• All-in on coding agents since at least Claude Opus 4.1 (August 2025). You should be using them daily and have strong intuition about what works and what does not
• Experience building harnesses, skills, or tooling around coding agents to improve their reliability and output quality
• Actively using more than one coding agent (Claude Code, Codex, Gemini, or others) and able to compare their strengths and weaknesses
• Demonstrated ability to iterate quickly, run experiments, and translate results into production improvements
• Clear, direct communication style. You write well and explain technical decisions to non-technical stakeholders

Preferred

• Experience building evaluation frameworks or benchmarks to measure AI system performance
• Deep understanding of software development lifecycle practices, including testing, CI/CD, code review, and deployment
• Background in developer tooling, internal platforms, or infrastructure that other engineers depend on

Not a Fit If

• You use coding agents casually but have not thought deeply about how to make them reliable and consistent at scale
• You are primarily a data scientist or ML researcher looking for a model training role
• You get attached to particular solutions or approaches. The state of the art moves fast and we need people who follow the research, run experiments, and go where the data leads rather than defending what worked last month

About Pit

Where intent becomes reliable systems

Our vision is a world where the friction between an idea and a production-ready system is zero.

Our mission is to help companies run their operations on custom software they can trust.

Our Principles

1. Start With the Customer

Everything we build serves someone. We stay close to real users and workflows to understand their constraints and outcomes, testing against reality rather than our own taste. Technical elegance is nice, but without user value it is redundant work. We do not build features to feel productive. We build them because they create clear, durable value for the people using what we make.

The customer is not a stakeholder to satisfy. They are the reason we exist.

2. Think Like an Owner

We take extreme ownership of our work and our mistakes. When pressure hits, we don’t point fingers or hide. We flag problems early, make a plan, and focus on the solution. We believe accountability is the fastest way to learn and improve.

Speak up. If something is broken, unclear, or heading in the wrong direction, say it directly, early, and to the right person. We hold strong opinions but remain willing to change our minds when presented with better arguments. Silence is not neutrality; it is a choice that lets problems compound. We do not punish people for raising hard things. We expect it.

3. Set the Standard

Quality compounds, but so does mediocrity. We hold exceptionally high standards in our product, code, design, and hiring. We do what’s right, even when nobody’s watching. We will take shortcuts, but conscious ones. We document them so they don’t become debt nobody owns.

The bar is not what we produce today. It is what we commit to produce every time.

4. Build Lego Blocks, Not Artworks

We do not build in silos. We maintain a big-picture view of the product, the architecture, and the company to ensure our decisions are scalable and durable. We choose the right level of complexity to solve root causes, not surface symptoms. “Good enough and fast” is a legitimate choice, but only when it makes the next thing easier to build, not harder.

That is the standard. Everything we ship should compound. Every decision we document should make the company smarter. Our job is not just to solve today’s problem well. It is to leave the platform, the culture, and the codebase better than we found them.

5. Win Together

No one carries the weight alone. When the load is heavy, we rally beyond role boundaries because we care about each other’s performance and well-being. High talent density only works if we operate as one unit.

Act with good intent. Assume your colleagues are trying their best. Give feedback to help, not to score points. Challenge decisions openly before they are final, then commit fully once they are. We do not play politics. We do not undermine quietly. We trust each other enough to be honest, and we expect the same in return.

6. Empower People with Technology

Technology is powerful and comes with real societal changes. We don't pretend otherwise and we take that seriously. Every solution we ship should make organizations and the people in them more capable, not just more efficient.

We think about where the human goes, not just where the workflow goes. Technology done right lifts the ceiling on what people can focus on.