• Sep 18, 2026LLM Inference on a Budget: What Actually Moves the NeedleKV-cache, quantization, batching, and speculative decoding — the practitioner shortlist for serving LLMs fast without a GPU cluster.
• Sep 10, 2026Faster Vision Inference: From Detector to Deployed PipelineONNX, quantization, resolution trade-offs, and tracking tricks — how I squeeze real-time perception out of Jetson and CPU edge boxes.

All posts →