Portfolio and blog: GenAI/agentic systems, distributed backend work, and notes on what I'm learning.https://swatisisodia.vercel.app/https://swatisisodia.vercel.app/blog/agentic-architectures-compared/https://swatisisodia.vercel.app/blog/agentic-architectures-compared/I built five ways to structure a multi-step AI agent and compared them on the same task. The design that re-plans after every step, the one that sounds smartest on paper, never once decided on its own that it was done, in 40 separate tries.Thu, 17 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/comparing-prompting-techniques-for-intent-classification/https://swatisisodia.vercel.app/blog/comparing-prompting-techniques-for-intent-classification/I compared seven prompting techniques on a 77 class classification task and measured both accuracy and operational cost. One technique won by a wide margin, and one made things worse.Tue, 08 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/context-window-management-compared/https://swatisisodia.vercel.app/blog/context-window-management-compared/I compared four ways to decide what actually stays in a model's context window when there's more content than comfortably fits. One technique failed as a clean step function, not a gradual decline. Another failed completely, 16 times out of 16, whenever the fact it needed got summarized away.Wed, 23 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/evaluation-criteria-across-llm-rag-and-agent-systems/https://swatisisodia.vercel.app/blog/evaluation-criteria-across-llm-rag-and-agent-systems/A framework for which evaluation metrics actually matter across four system shapes, plain LLM, RAG, single agent, multi-agent, then tested for real: the same task run through all four, where single-shot retrieval tied the winning multi-agent system while doing a fraction of the work.Fri, 18 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/peft-lora-qlora-a-theoretical-comparison/https://swatisisodia.vercel.app/blog/peft-lora-qlora-a-theoretical-comparison/I compared six ways to fine-tune a large language model by computing real trainable-parameter counts and memory footprints, no training run required. QLoRA's whole memory win turned out to be compression, not training fewer parameters.Wed, 09 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/rag-architectures-compared/https://swatisisodia.vercel.app/blog/rag-architectures-compared/I built six different ways to do retrieval-augmented generation and compared them on the same corpus and the same questions. The one that won overall wasn't the one that found the best evidence.Sun, 13 Sep 2026 00:00:00 GMThttps://swatisisodia.vercel.app/blog/transformer-deep-dive/https://swatisisodia.vercel.app/blog/transformer-deep-dive/A from-scratch walkthrough of the original transformer architecture, a worked attention example computed by hand, and how BERT, GPT, T5, and the efficiency and Mixture-of-Experts designs since 2017 changed it, and why.Wed, 09 Sep 2026 00:00:00 GMT