Blog posts

All tags
mDNS
AWDL
Networking
Distributed Systems
Jetson
LLM Inference
Benchmark
1-bit LLM
Cluster Setup
Edge AI
llama.cpp
CUDA
Energy Efficiency
NVIDIA Jetson
GRPO
Reinforcement Learning
LLM Training
Apple Silicon
ML Infrastructure
Raspberry Pi
Python
Edge Compute
Distributed Training
Thunderbolt

No posts match your filters.

2026

Clustering 3 Jetson Orin Nano Super

14 minute read

Build a 3-node Jetson Orin Nano Super 8GB cluster with active cooling. Real numbers: ~759 Mbps per link (gigabit), peak 58.3°C across all 3 nodes under full 18-core sustained load, zero throttling at 1728 MHz throughout. Read more

smoltorrent: Distributing ML Checkpoints Across a Pi Cluster

24 minute read

A 942 MB checkpoint. Four Raspberry Pis. ~1.5 min gather. No single point of failure. A deep dive into smoltorrent - a distributed checkpoint sharding system built over raw TCP with replication, SHA-256 integrity verification, mDNS discovery, and Prometheus monitoring. Read more

Clustering 4 Raspberry Pis 4B

14 minute read

Build a 4-node Raspberry Pi 4B cluster with UCTRONICS enclosure, PoE+ hats, and TP-Link LS110P PoE switch. Real numbers: 94.4 Mbps per link (100 Mbps switch ceiling), 62.3°C under full 16-core load, zero throttling at 1800 MHz throughout. Read more

Mac Minis Thunderbolt Cluster Setup Guide

9 minute read

Wire Mac minis into a high-bandwidth local Thunderbolt cluster for distributed training and inference with zero cloud egress cost, low latency, and direct control over cluster networking. Read more