Senior AI Engineer

Pixee•Baltimore, MD

2d•Hybrid

About The Position

Enterprises now generate more code than their security teams can review, and the backlog compounds every quarter. Pixee builds the harness for security engineering. The layer that catches risk at design, clears backlogs at machine speed, and validates every change before it merges. Dashboards don't reduce risk. Triaged findings and merged pull requests do. We're hiring a Senior AI Engineer to build the agents that do that work. You'll sit on the AI side of the house and both read and write a lot of Python. You'll build applications on top of AI coding agents that read enterprise codebases, propose real changes, verify them, and prove they're right often enough to merge.

Requirements

5+ years building production software, with deep Python expertise.
Real experience building agentic workflows, tool design, context engineering.
You've shipped something people depend on, not just prototyped.
Evidence that you measure your own work. You've built benchmarks, you know your system's failure modes, and you can state its accuracy without guessing.
Enough application security literacy to reason about the OWASP Top 10, why static analysis produces the false positives it does, and what makes a fix correct rather than merely green.
Comfort reading code you didn't write, in languages you don't primarily work in.
The ability to own a problem end to end and bring other engineers with you.
Based in the Baltimore, MD area, with the ability to be in office two days a week (typically Mondays and Wednesdays).

Nice To Haves

Static analysis internals — AST manipulation, dataflow analysis, or writing codemods.
Integration experience with the scanner ecosystem (Semgrep, CodeQL, Sonar, Snyk).
CI/CD and tooling that lives inside a developer's workflow.
Contributions to open-source AI or security projects.

Responsibilities

Build with agents, not just models. Ship applications where AI coding agents do the work of tracing execution paths, proposing fixes, opening pull requests. You own the tools, scaffolding, and control flow that make them reliable.
Move from detection to decision. Turn raw findings into exploit-aware judgments, and turn those judgments into changes a developer merges without arguing.
Make the context graph earn its keep. Decide what an agent needs to know (call graphs, execution paths, prior fixes, institutional history) and get it in front of the model at the moment it matters.
Prove it. Build and extend the evals that gate our work. Nothing ships on a good demo. It ships on a measured merge rate against real repositories.
Run it in production. Own the performance, cost, and reliability of systems running continuously across thousands of enterprise repositories.
Pair with customers. Work with customers and prospects to setup, configure, and integrate Pixee into their workflows and infrastructure
Raise the floor. Review thoughtfully, write code others can pick up, and mentor the engineers around you.