Job description
About Us At SkyLabs AI Inc.
, we are at the forefront of the artificial intelligence revolution.
As a US-headquartered startup, we are building verification infrastructure for C++.
We move fast, we ship constantly, and the people who thrive here are the ones who don't wait to be told what to build next.
The Role We need someone who deeply understands C++, compilers, and systems, but who builds the tooling in whatever language is fastest: Python, TypeScript, Rust, shell, glue code, prototypes.
You'll be expected to identify what needs to exist, build it, and put it in front of users without a project manager holding your hand.
This role is unapologetically AI-native.
We expect you to drive Claude Code, Codex, Cursor, or similar agentic coding tools as a core part of your workflow, not as a novelty, but as the way you ship.
Who thrives here People who treat a vague problem statement as a starting point, not a blocker People who default to action: build the prototype, run the experiment, ship the first version, then iterate People who hold a high bar for themselves without needing anyone to enforce it People who are energized by ambiguity and a fast cadence, not exhausted by them We don't require prior Python or TypeScript experience.
If you have strong C++ and systems judgment and you know how to drive a coding agent, you can produce tooling in whatever language the job needs.
If you need fully-scoped tickets, weekly 1:1s to unblock you, or a stable roadmap that doesn't shift, this is not the right environment.
Competitive salary in USD (tax-free) Comprehensive health allowance Relocation allowance (if you're moving to Islamabad) Monthly team events and offsites A beautiful, collaborative office space Work alongside world-class AI researchers and engineers What you'll actually do Build developer tooling around a C++ codebase: code generators, AST analyzers, build orchestration, test harnesses, fuzzing infrastructure Work with Clang/LLVM where it matters (AST traversal, instrumentation, code generation), but reach for Python or TypeScript when that's the right tool Design property-based testing infrastructure and translate formal specifications into executable runtime checks Set up and tune fuzzing pipelines (libFuzzer, FuzzTest) Decide what to build, build it, ship it, learn from it, repeat You should have Strong working knowledge of systems programming and modern C++ (C++17/20, ideally some C++23/26 contracts exposure), enough to read large codebases, reason about ABI, performance, and build systems, and understand what tooling needs to do to C++ code Compiler/systems intuition: ASTs, IRs, linkers, sanitizers, build graphs.
You don't need to have written an LLVM pass, but you should understand what one does Build systems experience: Make, CMake, or Bazel on non-trivial projects Demonstrated use of agentic coding tools (Claude Code, Codex, Cursor, etc.
) on real work.
Be ready to talk about what you've shipped with them, where they helped, and where they failed you A working mental model of LLMs: context windows, tool use, prompt engineering, where models are reliable and where they aren't A track record of operating independently in fast-moving environments.
Previous startup experience is a strong signal Bonus points Fuzzing experience (libFuzzer, AFL++, FuzzTest) Property-based testing (QuickCheck, Hypothesis, FuzzTest, RapidCheck, etc.
) Clang/LLVM internals: libTooling, LibClang, plugins, sanitizer internals C++26 contracts knowledge or design-by-contract background Exposure to formal methods, SMT solvers (Z3, CVC5), or proof assistants (Lean, Rocq, Isabelle) Open-source contributions to compiler/language tooling Not a fit if You want to write only C++ all day.
Most of the tooling could be another language You are completely unwilling to use AI coding tools.
Being aware of their limits and healthy skepticism is fine You need fully-specified tickets, predictable scope, or a stable roadmap You wait for permission before building
This job post has been translated by AI and may contain minor differences or errors.
