February 2026 monthly summary for fastmachinelearning/hls4ml. Key features delivered: Keras v3 model conversion and numerical profiling compatibility fix enabling accurate profiling for Keras v3 models, with added tests to cover Keras v3 paths and resolution of dependency namespace conflicts. Major bugs fixed: Fix Keras v3 model conversion in numerical profiling; ensure input model readiness before profiling; resolve qkeras/hgq2 namespace conflicts and align profiling tests (bar counts and data initialization). CI and dependency hygiene: updated CI YAML/test lists for Keras3 paths, reintroduced hgq2 where needed, and applied pre-commit auto-updates to improve maintainability. Overall impact and accomplishments: Broader, more reliable Keras v3 support in hls4ml profiling, leading to fewer breakages in production workflows and smoother migration for users with modern Keras versions. Technologies/skills demonstrated: Python-based profiling tooling, Keras v3 compatibility, oneAPI elements (Einsum/EinsumDense), test-driven development, CI automation, dependency management, and pre-commit tooling.