Aditya
ExperienceProjectsResearchBlog

Blog

Writing about ML systems, fairness in AI, and the occasional space tangent.

Fine-Tuning a Small LLM with DPO for Less Than 50 Cents: What Actually Happens

A hands-on account of building a DPO fine-tuning pipeline from scratch on a $0.44/hr GPU. The seven errors you will actually hit, the LLM-as-judge evaluation, and the one config change that made training twice as fast.

April 18, 202610 min read

The Car in My Head: A No-Nonsense Tour of the LLM Words Everyone Throws Around

SFT, DPO, LoRA, vLLM, agents, the harness. Every LLM term mapped onto one mental picture: a car. Once you can tell training from serving from using at a glance, the jargon stops sounding foreign.

April 2, 202612 min read

© 2026 Aditya Mamidi · Santa Clara, CA

GitHubLinkedInXScholar