Jobs at Lemurian Labs

Build and maintain the front end of a portable AI compiler: parsing, validation, and IR lowering from frameworks (PyTorch, StableHLO, ONNX, MLIR) into internal IR. Ensure operator coverage, collaborate with optimization and codegen teams, and improve correctness, tooling, and ingestion pipelines.

Design, build, and maintain a multi-target runtime for an AI compiler stack. Implement low-level parallelization, kernel scheduling, and performance analysis; prototype runtime ideas; benchmark compiler outputs on target hardware; build tooling to collect and analyze performance bottlenecks; collaborate with compiler and product teams to improve runtime architecture for diverse hardware and large-scale execution.

Design and implement code generation features for a heterogeneous AI compiler, optimize low-level target-specific code, apply parallelization and kernel generation techniques, use performance data to guide improvements, and collaborate with product teams to align compiler architecture with ML needs.

Design and build a performance testing platform for LLM inference on GPU clusters. Define benchmarks and metrics, run automated validation pipelines, profile GPU workloads, establish baselines, validate compiler optimizations, and deliver dashboards and documentation while collaborating with compiler, ML, and DevOps teams.

Lead developer adoption of the company's AI/HPC stack: design the developer journey, act as Customer Zero to feed market insights into the product backlog, craft technical narratives and documentation, cultivate open-source and cloud-native ecosystem relationships, and define/track adoption metrics with growth engineering.

Founding member of the Developer Experience team building profiling, debugging, and performance-introspection tools for a compiler stack. Create client-side tooling to surface server-side compiler telemetry, expose hardware performance and interrupt data, develop GPU debugging visualizations, define debug/data-exchange formats, and collaborate with engineering teams toward internal and external tooling.

Design, implement, and maintain graph-level optimization passes (fusion, layout, dead code elimination, constant folding). Evolve IR, analyze performance to improve throughput and latency, collaborate with front-end and backend teams, prototype new optimization strategies, and contribute to testing/validation to ensure correctness across models and hardware.