Economist. Chief of Staff at a US B2B consulting firm. PhD candidate in Economics at Korea University Sejong Campus, working on Brazilian higher education with national administrative microdata.

I work on the same question in two places: what people and organisations do when incentives and institutions change. In research that means panel data, difference in differences and survival models. In practice it means the forecast, the board materials and the commercial models behind a decision. Everything below is built in that overlap.

Brazilian, based in Sejong, South Korea. Portuguese, English, Spanish, and enough Korean to get by.

BreakTalk, on Chief of Staff practice, AI adoption, and the economics underneath both.

My CV and full record, published at ingridleiria.github.io.

More than 100 open working instructions for Chief of Staff work and empirical research, MIT licensed, in five tracks: Chief of Staff, PhD research, brand and writing, commercial and data, and web and presentation. They are plain Markdown methods rather than prompts, so they run in Claude, ChatGPT, Gemini or a coding assistant. The argument behind the library is that a clever prompt is tied to one model and a written method is not. The whole catalog is published as data, every skill is checked against the same standard on every push, and there is a run from end to end showing what a written method does that a good prompt does not.

Daily price, quality and liquidity intelligence for the secondhand luxury handbag market in South Korea and Brazil. Ten authenticated resale platforms are scraped every day and written into a longitudinal panel, one observation per listing per day, which makes price trajectories and time to sale reconstructable. That panel feeds hedonic price models, value retention, and survival models of how quickly a bag actually sells. Trilingual product on React, Node.js and PostgreSQL with pgvector.

ECONOS, an AI econometrics workspace. Describe a model in plain language, get working Stata, R or Python code back, run it in an integrated terminal, and export output formatted to replication standards. Built because students with good economic intuition were being stopped by syntax rather than by economics, and it carries a gamified path that teaches the same methods. The architecture note explains why generation and verification are kept structurally apart, which is the decision the whole system rests on.

VIZZ, a WhatsApp assistant in production at a private surgical practice in southern Brazil, built with a physician collaborator. It handles the ordinary parts of patient communication in Portuguese, and it is designed around the moment it should stop: when a message needs clinical judgment, or the model's confidence is not high enough, the conversation goes to the practice rather than to an improvised answer. The architecture note argues for making that boundary a layer of its own rather than a paragraph in a prompt, and how this was built is the honest account of putting a system into a regulated setting without an engineering team.

Applied microeconometrics: Brazilian higher education, illegal markets, entrepreneurial entry. Three research prizes from Brazil's federal and regional economics councils between 2016 and 2018, for work on decision making under economic uncertainty and on illegal markets. Publications and the full record are on my page and on ORCID.

LinkedIn · BreakTalk · ORCID · ingridleiria.github.io · ingrid@leiriaconsulting.com