I'm Adi, an Applied Scientist at Amazon and a former AWS Bedrock engineer. I work on agentic AI, Model Context Protocol (MCP), and RAG systems. I also teach an Agentic AI workshop every year at NYU.
Check out the link below to see some of our content!
I's till you're 18
Ahaan was 12 when he first met you. A round faced reality show
contestant dealing with fame and insecurities.
Ahaan was 18 when he last met you. A pimple-faced adult who had
dealt with some large demons.
He was your typical teenager - but with the bizarre need to document
everything.
His life was ordinary in a lot of ways. Heart breaks, betrayals,
regrets, fears, musings... Everything that the world talks about.
Yet, he had had some extraordinary experiences - ones which the
world hardly ever talks about.
He captured every phase of his life with poems and short stories,
which led to the birth of this collection - a 6-year conversation
between you and Ahaan,
where you will see the world through the eyes of a boy -
through the I's of a boy -
right up till the age of 18.
My work spans production RAG at AWS Bedrock, MCP-based engineering automation,
and the evaluation of AI agents. Here are some public tools and writing from that work.
I also teach an agentic AI workshop at NYU,
with recordings and code you can explore.
Open-source & tools
Developer tools and research I'm building in the open.
~/human-hindsight
▸The prompt you'd write if you knew then what you know now.
Turn your Claude Code, Codex, Kiro, and OpenCode history into reusable first prompts
and working preferences for CLAUDE.md, AGENTS.md, or Kiro steering.
Start with a guided CLI, then browse, search, and copy prompts in a local Markdown
library. Scanning and export stay offline; distillation uses your chosen agent CLI
after you approve the model calls.
▸A scientific loop for making LLM prompts measurably better.
Scientific framework for iterative LLM prompt improvement using multi-dimensional scoring,
threshold optimization, cross-validation, and an OPRO-style agent loop.
