FlagOS TLE Kernel Developer
Develops, optimizes, and debugs high-performance TLE GPU kernels and compiler lowering pipelines with rigorous validation workflows.
The FlagOS TLE Kernel Developer skill orchestrates end-to-end development for Triton-like Extension (TLE) kernels and FlagTree compiler features. It guides engineers through the full lifecycle of custom hardware acceleration—from intake and implementation through reproducible verification, performance profiling, and merge decision packaging. Designed for deep learning engineers and compiler developers, this skill ensures consistent validation guardrails, structured fix summaries, and systematic performance tuning when building TLE APIs, verifiers, lowering passes, or specialized deep learning operators.
Características Principales
01Orchestrates end-to-end TLE kernel development from task intake to merge decisions
02Guides optimization and profiling for high-throughput AI and ML operators
03Generates standardized fix summaries, regression records, and post-mortem lessons
0422 GitHub stars
05Enforces strict reproducible validation matrices and non-negotiable code guardrails
06Provides structured workflows for TLE compiler lowering, API, and verifier features
Casos de Uso
01Implementing and testing compiler lowering passes and verifier features in FlagTree/TLE
02Writing and tuning custom high-performance TLE operators for neural network layers
03Debugging memory safety, correctness regressions, and pointer issues like local_ptr