AI product manager by day, builder the rest of the time. I ship AI products end to end: the reasoning, the interface, payments, deploy. Building them is how I stay current on what actually holds up in production: which model to use where, what breaks, what costs too much. That judgment is what I bring to the products I ship at work. Here's the running list, with a link to try each one.
Scores a startup idea on market size, competition, and real demand in about a minute, so you know before you build.
My proof I can take an AI product the whole distance: reasoning pipeline, UI, payments, deploy. Open source, and people pay 5 dollars a report. Same shape as the hard enterprise problem, pull noisy real-world signals and hand back a verdict someone trusts.
A knowledge base with AI-native recall, free for anyone to run. Notes, links, documents, and tasks are embedded the moment you save them, so you can search by meaning and ask questions across everything you have kept. YouTube links come back with the transcript summarized, screenshots are described and made findable, and a wiki layer folds each new capture into a topic page so the archive compounds instead of piling up.
The most useful thing here for someone who is not me. MIT licensed, self-hostable in a few minutes, and it ships an MCP server, so any AI conversation can search and write to your own brain. It is also the memory the rest of my agents draw on.
AI product photography for apparel. A phone flat-lay of a garment goes in, a clean on-model studio shot and styled scenes come out, at around five cents an image against roughly a dollar from the incumbents.
The build I am running in the open right now, costs published as they land. Every model vendor sits behind a one-function swap point, so a cheaper or better image model is a one-line change instead of a rewrite. Picking which dependency will churn is the product work, and it is the part that outlives the model.
A lightweight Notion alternative with kanban boards, a rich-text editor, and workspace collaboration.
Collaboration tools get heavy because every team asks for one more field. This was the test of how much you can cut and still have something a team will use. Open source, so you can cut further.
Reads a code repo's legal and security signals to flag whether it's safe to install, fork, or depend on, without running it.
Supply-chain trust, grounded in evidence instead of a gut call.
A full web presence for a Hyderabad interior-design studio: portfolio, design philosophy, and a structured inquiry flow. Behind it sits a quotation builder and a product catalog I built to run the operations side.
The whole product, not only the brochure. A real studio runs its day on the internal tools, which is a harder brief than a marketing site and a better test of whether the thing works.
An app I built for my wife during her pregnancy. Science-backed reminders for nutrition, hydration, and medication, personalized to her routine, preferences, and daily inputs.
Built for one person who matters, which is the most honest test of a product. Health guidance has to be trustworthy and personal, so it is.
A health leaderboard where friends compete on staying consistent and strong.
Behavior change is a product problem, not a willpower problem. This tests whether a leaderboard can drive it.
A multi-agent system I built to run my own job hunt. One source of truth for my experience, then it tailors a resume to any role, scores the fit, researches the company, and preps the interview. Reviewer sub-agents grade the output, and daily job briefings run on a schedule.
The honest way to prove I can design an agentic system is to live on one. Roughly 18 skills plus sub-agents, orchestrated and version controlled. I used it to prepare for this conversation.
Turns my notes and point of view into LinkedIn posts. Same engine as the X agent, tuned for a different room.
Presence without the time sink. Drafting is automated, the final cut stays mine.
My X account, where I build in public: what one person and a few agents can actually ship, with the costs and the failures left in. A content agent I built drafts in my voice and to my editorial standards.
Working in the open keeps me current and keeps me honest. The agent handles the mechanical part, I keep the judgment.