Systems & Hardware

Scaffold a custom Jetson carrier board overlay from a reference devkit

Snapshot Jetson health across memory, GPU, thermals, power, storage and processes

Download Jetson Linux BSP tarballs, sample rootfs and public sources

Flash a promoted BSP image onto a Jetson in recovery mode

Generate a per-target knowledge base from a Jetson BSP source tree

Plan and apply headless mode on Jetson to free GUI and daemon memory

Extract Jetson Linux and sample rootfs tarballs and apply the binaries

Set up a Jetson BSP source workspace with a Crosstool-NG toolchain

Author a new Jetson target profile for a reference devkit and carrier

Link pre-downloaded Jetson reference docs into the active target profile

Measure Jetson DRAM and NvMap usage before and after memory reclaim

Reclaim Jetson DRAM by disabling unused subsystems in boot and kernel config

Print L4T version, board configs and rootfs state from a BSP tree

Print module model, L4T version, kernel and power mode from a running Jetson

Promote overlay files and built artifacts into the staged BSP image

Start Jetson or IGX BSP customization with a guided setup questionnaire

Switch the active Jetson target profile before a build or flash

Run static checks and on-device smoke tests on a flashed Jetson image

Check drivers, tune games and set laptop features through the NVIDIA App

Collect and package NVSHMEM bandwidth and latency results with topology evidence

Map NVSHMEM NICs to processing elements and export matching environment variables

Prepare NVSHMEM CUDA kernels for TMA shared memory registration and transfers

Learn NVSHMEM with a tutorial covering first C, C++ and Python programs

Plan and check NVSHMEM and NVSHMEM4Py installs from packages, containers or source

Pick an NVSHMEM remote transport for inter-node runs from system evidence

Diagnose NVSHMEM launch failures, hangs and transport issues and draft bug reports

Route NVSHMEM performance tuning to the right data, transport or mapping skill

Choose a robot simulation physics engine by comparing Newton, PhysX, MuJoCo and Drake

Check and install compatible NVIDIA drivers, CUDA and container toolkit on GPU hosts

Add a cuTile GPU kernel to TileGym with dispatch registration, tests and benchmarks

Port cuTile Python GPU kernels to cuTile.jl, handling indexing and layout changes

Convert cuTile GPU kernels to Triton and debug illegal address and shape errors

Add or debug CuTile autotuning with cached tune-once and launch patterns

Write validated high-performance GPU kernels with the cuTile tile programming model

Profile and tune cuTile kernel speed across tile sizes, occupancy and autotune

Reduce NVIDIA Warp compile and startup time, including JIT recompiles in CI

Judge whether a slow code path is a good candidate to port to NVIDIA Warp

Build Go command-line tools with flags, config layers, exit codes and shell completion

Build Go CLI command trees with cobra, flags, argument checks and completion

Load Go app config from flags, env vars and files with viper precedence

Find disk space, memory hogs and startup bloat on macOS, Linux and Windows

Check for the uv Python package manager and install it if missing

Drive tmux sessions by sending keystrokes and reading pane output for interactive CLIs

Run Python scripts and manage dependencies with uv instead of pip and venv

Read sensors and send commands to cameras, lamps and arms through hardware adapters

Capture and filter Windows debug output with Sysinternals DebugView from the CLI

Structure Dart command line apps with args, exit codes and subprocesses

Compile C and C++ sources into Dart native assets with build hooks

Generate Dart FFI bindings from C headers with ffigen instead of by hand