Software Engineer III, Copilot Agents GitHub

About Git Hub

About Git Hub Git Hub is the world’s leading platform for agentic software development — powered by Copilot to build, scale, and deliver secure software. Over 180 million developers, including more than 90% of the Fortune 100 companies, use Git Hub to collaborate, and more than 77,000 organisations have adopted Git Hub Copilot.

In this role you can work from Remote, United Kingdom

Git Hub is the home for software development, where we collaborate to build the world’s leading AI-powered developer platform.

In the Copilot Agents organization at Git Hub, we are passionate about ensuring the security and quality of the world’s software – from open source to the enterprise, written by humans and by AI tools. We believe that the best way to secure and improve the quality of software is to detect actionable issues early in the development process, and actively facilitate their remediation as part of the developer workflow.

Our team develops detection and remediation engines that power several Git Hub products used by hundreds of thousands of developers and projects every day:

• Copilot Code Review:
  Git Hub LLM-powered code review engine which provides feedback for immense volumes of code changes daily helping developers fix bugs and improve the quality of their code, leading to faster time to merge.
• Copilot Autofix:
  Git Hub’s LLM-powered remediation engine that produces high-quality fix suggestions for security or quality findings, empowering developers to fix them as soon as they are found or burn down the debt already existing in their codebase. It is used as the remediation engine within the Git Hub Code Security, Git Hub Advanced Security, and Copilot code review products.
• CodeQL:
  Git Hub’s semantic code analysis engine that uses world-class static analysis research and technology to deeply analyze code, enabling the early detection of security vulnerabilities and correctness errors in software. CodeQL supports a wide range of programming languages, including C/C++, C#, Go, Java, JavaScript/Type Script, Kotlin, Python, Ruby, and Swift. It is used as the primary detection engine within the Git Hub Code Security and Git Hub Advanced Security products.

We work as a distributed group within a distributed company. The majority of our team members live across Europe, the US, and Canada, and while we do have some offices, all our meetings are location-agnostic and happen online. We operate with a high degree of autonomy and trust, and we have a significant level of influence on the product and technical direction of security and code review products  value learning, introspection and reflection, and we’re always looking for ways to improve as a team and as individuals, so candor and a culture that values safety to speak up are highly important to us.

We are looking for a software engineer to join one of the distributed software engineering teams responsible for building and expanding code analysis engines and agents  this role you will work in one of the following areas, selected based on your own experience and interests, and the needs of the organization at the time of hiring:

• Code analysis
  :
  Maintaining detection support for multiple programming languages, including:
• building source code extractors that translate code written in each language into data that CodeQL can understand, keeping up with the latest version of each supported language
• writing and maintaining queries in the CodeQL query language that accurately detect security vulnerabilities and undesirable coding patterns
• ensuring the quality of LLM-powered Copilot Autofix suggestions for fixing the alerts found
• building analysis libraries in the CodeQL query language that reason precisely about the semantics of programming languages, supporting the development of queries across all languages
• experimenting with and robustly evaluating LLM-powered detection engines and integration between LLMs and traditional static analysis.
• AI agents:
• LLM-based code review:
  Building, improving, and evaluating the detection agent for Copilot code review, using LLMs and contextual information from a codebase to produce AI-powered…