Software Engineer

Software Principal Engineer – Autonomous AI Software Engineering

Software engineering is being reshaped by autonomous AI agents that can plan, write, test, and review code with increasing independence. Our team works at the forefront of that shift, bringing agent-driven autonomy into the software development lifecycle so that engineers set direction and make key decisions while agents handle more execution. 

Because the agent landscape evolves rapidly, we stay hands-on: exploring the most promising approaches, proving them with rigorous experiments on real engineering work, and turning successful patterns into standard development practices. We are a small, AI-native team operating like a startup—prioritizing high autonomy, rapid experimentation, and daily human-AI collaboration.

Join us to do the best work of your career and make a profound impact as a Software Principal Engineer on our AI Governance Engineering team in Hopkinton, Massachusetts. This is an onsite role.

What You'll Achieve

You will be the hands-on engineer who turns autonomous agents from impressive demos into dependable engineering practice. You will set up and run proofs of concept of agent platforms and approaches against real codebases and tasks, measure objectively what they can and cannot do, and integrate viable solutions into our development lifecycle—operating within the architecture and evaluation framework defined by our architect. 

You will spend most of your time building and experimenting (70%+), using AI agents as a core part of your daily workflow.

Responsibilities

Key Responsibilities:

• Run hands-on proofs of concept and pilots of autonomous software-engineering agents and agent platforms, from environment setup and integration through results, on representative real-world engineering work
• Build evaluation harnesses: realistic task suites, benchmarks, and metrics for output quality, level of autonomy, reliability, speed, cost, and security, enabling fair and repeatable comparisons
• Verify and validate agent output using automated tests, mutation testing, static and security analysis, acceptance-criteria checks, and calibrated human review to identify failure points
• Integrate agents into the engineering toolchain (source control, CI/CD, issue tracking, specifications, and quality gates) via APIs, MCP, and custom tools, building extensions where standard solutions fall short
• Increase autonomy safely by implementing guardrails, sandboxing, observability (traces and replay), and feedback loops, helping engineering teams adopt agent-driven workflows with confidence