I’m Luba, a software engineer in New York. I spent two years at Twitter on the core Tweet infrastructure, and now I build AI-powered products from the first conversation with a user to the deploy.
I’m picky about where AI goes. Rules people depend on stay in plain code. AI gets the jobs where a person checks the result.
A four-person office gets restaurants their liquor licenses. Every application, stage, deadline and document used to live in spreadsheets, email and memory. Now it lives in one place.
Read the case studyA daily research agent for maternal health software companies. It finds hospitals whose postpartum outcomes don’t match their public promises, scores the gap and drafts the outreach.
Read the case studyShows a venture firm how close it is to each person in its network, and why. I built the database, the data enrichment and the MCP server.
Read the case studyTells a Brooklyn bar owner how many drinks they need to sell each day to cover rent, staff and everything else. It shows how price changes move the target, lets you model a weekend-heavy week, and compares your costs to neighborhood averages. Numbers stay on your device.
Try it liveA web version of Apple Notes with folders, search and the familiar three-column layout. Notes save on their own 800 ms after you stop typing and again when you click away. Every save carries a revision number, so a slow, older save can never overwrite newer edits.
See it on GitHubFor Honcho (Plastic Labs), a memory layer for AI agents, I built a skill with three tools (save, query and get context), plus tests and setup docs; it was reviewed and merged. For Cal.com, I opened an accessibility fix for booking-calendar labels and contrast.
I came to software from nutrition. I still work the way I learned there: start with the person, figure out what they actually need, then build the thing that fits.
I’m originally from Russia and have lived in New York since 2005, and I speak English and Russian. I studied nutritional science, managed recipe and ingredient data at Plated, and then changed careers through Pursuit.
At Twitter I worked on the Tweet Foundation team, on code every feature team in the iOS app depended on, and led a big modularization milestone there.
I’m comfortable owning a product end to end and just as comfortable as part of a team. At Pursuit I lead project teams that change every few weeks, so I’m used to getting new people moving quickly.
A lot of my work starts vague: a process nobody has written down, a problem the client can’t quite name yet, requirements that shift after the first demo. I’m at home there. I also care about what comes after, the structure and rules that keep a system reliable once people depend on it. That’s what I learned at Twitter.