Erfan
Boostan
I build and ship complete AI products. I founded Team Memory (teammemory.online), an AI that tracks the decisions and promises in a team's chats across Telegram, WhatsApp, Slack and Discord, built on Next.js, Supabase and the Claude API. I also shipped KibRide, a ride-hailing app live on the App Store and Google Play, and built the frontend for Automech, a mechanic marketplace used in six countries. I work by directing AI coding tools, and I test what they give me before trusting it.
Experience
I built and launched an AI that takes notes in a team's group chats on Telegram, WhatsApp, Slack and Discord. It finds decisions, promises, ideas and open questions, links each one to the message it came from, and reminds the right person before something slips: in the group, privately, or in the owner's weekly roundup, and never at night. Its assistant works from the user's own WhatsApp or Telegram: with Plans, the user says who does what by when, Claude writes the messages, the user approves them before they go out, and it follows up until every task is done. I designed it to be trusted with private chats: it announces itself in every chat, anyone can hide a message, each team's messages are encrypted with their own key, raw chats are deleted after 90 days, and it tracks topics, never individual productivity. Next.js, Supabase and the Claude API, with a 3D memory map, a Claude connector and 401 automated tests. teammemory.online
I built KibRide, a ride-hailing app for North Cyprus, and shipped it to the App Store and Google Play (com.kibride.passenger.android). Kotlin Multiplatform for the code shared between Android and iOS, Firebase for the backend — sign-in, the database, and live ride tracking. Separate apps for passengers and drivers, plus the website. I wrote nearly all of it, between January and May 2026.
A marketplace for finding mechanics, used in Cyprus, Greece, Estonia, Latvia, Lithuania and Germany (app.automateonline.user). I built the website and designed the interface for both apps — the one customers use and the one mechanics use. React, Next.js, Tailwind.
Found my own clients and handled the negotiations myself.
Projects
An AI that answers a small business's customers by itself, in Turkish, English, Russian and Persian, using only that business's own stock and prices. One version for car galleries, one for estate agents. When a customer is serious, it saves them as a lead with a summary for the owner, then keeps reminding the owner until someone calls back. Python and the Claude API (tool use, prompt caching). I built it by directing Claude Code. I tested the AI against 12 customer scenarios in four languages, including a low cash offer and a prompt-injection attempt: 11 passed first time and I fixed the one that failed. Plus 62 automated checks and a browser test. github.com/er369ai/ai-customer-assistant
My own project — an app for meeting people and swapping skills in Cyprus. Two apps, one for everyday users and one for professionals, built in Expo/React Native on Supabase, plus the website. The idea, the design and the code are all mine. Not published yet.
A fixed-price, one-week website service for small businesses in North Cyprus. To find clients for it, I built a Python outreach tool. It collects businesses from listing sites, cleans up the contact details and checks my own WhatsApp history so nobody gets messaged twice (about 3,400 businesses so far). It drafts a message for each business with the Claude API, based on what that business sells, and I review and send every one myself. kktcweb-er369ais-projects.vercel.app