AI · Product · First principles
I'm Piyush. I build AI products and spend a lot of time understanding how the technology actually works — from the research behind models to the systems that make them useful.
I've spent my career bringing software products to life. These days, I'm exploring model training, inference, agents, and evaluation while building and sharing what I learn along the way.
Ideas and explanations I'm working through, from foundational research to practical AI systems.
Starting at the beginning: how a simple artificial neuron learns.
Draft in progressWhat it means for a neural network to learn from its mistakes.
Draft in progressFrom tokens and GPU memory to KV cache and generated output.
PlannedExperiments and tools that help turn complex ideas into something tangible.
Visual, hands-on explanations of landmark research papers — built to help curious learners develop intuition.
In progressExploring the full journey from preparing a dataset to training, fine-tuning, and evaluating a model.
ExplorationThinking through how to design dependable agentic workflows and measure whether they work.
Notes & experimentsSmall experiments at the intersection of intelligent systems and real, everyday problems.
ExplorationI'm a product builder who has spent much of his career shaping software products and working with engineering teams to bring them to market. Most recently, I've worked on enterprise AI platforms spanning data, models, agents, and evaluation.
What interests me most is connecting the fundamentals of AI research to useful products — and making those ideas understandable to more people. This site is a record of what I'm learning, trying, and building.