Long-form notes on the systems I build — machine learning, backend engineering, and the messy parts in between.
A personal account of building a production ML system from scratch, including the messy parts: weak supervision, fine-tuning on 4GB of VRAM, and deploying to a t3.micro.