I’m Sanjay (or heissanjay online)
I spend an unreasonable amount of time asking “why?” until abstractions give up and reveal what’s underneath. My hobbies include dismantling machine learning papers into first principles, teaching GPUs to stay busy, and convincing myself that implementing something from scratch is faster than reading the documentation. It rarely is.
By profession, I’m a ml engineer. By temperament, I’m an engineer who distrusts magic. If an algorithm works, I want to know why. If it scales, I want to know why. If someone says “AI just figured it out,” I become mildly unbearable.
Lately I’ve been wandering through the strange neighborhoods of machine learning systems, LLMs, recommendation engines, graph neural networks, distributed training, and GPU optimization. Every topic eventually ends with me drawing boxes, arrows, matrices, or discovering that the bottleneck was memory bandwidth all along. It usually is.
This digital garden is where I leave breadcrumbs for my future self.
Welcome to the garden. Things are always under construction, much like the human understanding of artificial intelligence.