1 Minute of Football, Analyzed in 7 Seconds
From a fragile rule-based system to a production ML pipeline that processes 1 minute of 720p drill footage in 7 seconds. TensorRT optimization, $0/mo training costs, and a full Apache 2.0 migration.
TensorRT Optimization Guide for Object Detection
A practical guide to converting YOLO and RF-DETR models to TensorRT engines. FP16 precision, batch optimization, and CUDA memory management for maximum throughput.
RF-DETR vs YOLOv11: When to Use What
An honest comparison based on real production deployments. Speed, accuracy, training requirements, and the specific scenarios where each model excels.
Why You Don't Need A100s for Training
How I cut training costs from $2,500/mo to $0 using L4 GPUs. Efficient data loading, gradient accumulation, mixed-precision training, and smart scheduling.
AutoML for Model Selection: DEHB vs PriorBand
Findings from my research at the AutoML lab. When to use dynamic ensemble HPO vs prior-guided optimization, and how to build proxy search spaces.
Want to know when new posts drop?
Follow me on LinkedIn where I share shorter-form technical insights weekly.
Follow on LinkedIn