Skill UI

Expert in developing complete, high-performance firmware and peripheral drivers for ARM Cortex-M platforms (STM32, Teensy, nRF52). Specializes in low-level hardware interaction, including implementing drivers for I²C, SPI, UART, and managing complex concurrency patterns using FreeRTOS/Zephyr. Provides critical guidance on optimizing for determinism, cache coherency, DMA transfers, and memory barriers to ensure reliable embedded system operation.

A comprehensive guide detailing the entire workflow for analyzing embedded firmware. It covers hardware-level extraction methods (UART, JTAG, SPI dump), filesystem analysis (SquashFS, JFFS2), binary reverse engineering using tools like Ghidra, and identifying common vulnerabilities (Buffer Overflow, Command Injection). Ideal for security researchers and embedded systems engineers.

This guide details the use of the GGUF format and quantization techniques for running large language models (LLMs) efficiently. GGUF enables optimal inference on consumer hardware, Apple Silicon, and systems requiring CPU-only operation. By compressing models using various K-quant methods (like Q4_K_M), developers can significantly reduce memory footprint and hardware requirements while maintaining high performance.

A comprehensive guide to optimizing Granola's output quality across three dimensions: audio transcription accuracy, AI enhancement quality, and overall processing speed. Learn best practices for hardware setup, ideal meeting habits, and structuring your templates to maximize AI output reliability and efficiency.

Mamba is a selective state-space model architecture delivering O(n) inference for million-token sequences, offering hardware-aware Mamba-1 and Mamba-2 variants that beat Transformers on speed without KV cache overhead.

Compress large language models quickly using Wanda, SparseGPT, and N:M pruning to enforce hardware-friendly sparsity, enabling faster inference and deployment without retraining.

This skill automates the setup and configuration of Ollama, allowing users to deploy Large Language Models (LLMs) locally. It performs system assessment (OS, GPU, RAM), selects optimal models based on hardware availability, and handles installation across macOS, Linux, and Docker. Use this when you require free, self-hosted AI alternatives, need offline inference capabilities, or wish to eliminate reliance on costly hosted API services.

A comprehensive guide demonstrating the full lifecycle of GPU resource utilization on Vast.ai. This includes searching for available hardware using CLI and REST APIs, provisioning a dedicated GPU instance (e.g., using PyTorch images), connecting via SSH, running deep learning workloads, and finally, securely destroying the instance to stop billing. Ideal for testing cloud integration or managing ML compute pipelines.

This comprehensive runbook provides rapid, structured procedures for incident response on Vast.ai cloud GPU instances. It covers common failure scenarios like spot preemption, training job crashes, and GPU hardware issues. Users can follow the steps for triage, mitigation, recovery, and postmortem analysis to minimize downtime and ensure computational continuity.

This comprehensive guide outlines security best practices for utilizing Vast.ai's cloud environment. It covers secure API key management, hardening SSH access to GPU instances, implementing proper data cleanup before instance destruction, and adhering to robust network security protocols. Use this when setting up, auditing, or deploying sensitive AI/ML workloads on shared cloud hardware.

Cirq is Google Quantum AI’s Python framework for building, simulating, and deploying noise-aware quantum circuits on both simulators and real hardware, complete with experiment templates, transformation utilities, and hardware-specific toolkits.

Hardware-agnostic quantum ML framework enabling gradient-based training of quantum circuits, hybrid quantum-classical models, and seamless integration with PyTorch/JAX/TensorFlow alongside device portability across IBM/Google/Rigetti/IonQ.