FounderProduct EngineerAI EngineerFull-Stack
I build AI-native software: one product, two kinds of users. People get an interface they enjoy, their agents get APIs and MCP tools they can actually call, and I stay with it long after launch.
More than half a decade in, I build platforms that hold up at scale: NoteBot alone has served over 4.7 million requests, and I've been the only engineer on systems a whole company runs on.
Open for consulting and advisory work. Get in touch
Started BUTEX NoteBot in 2018 and still run it. I take products from the first conversation to launch, and keep owning them.
Scope with the team, cut to what matters, ship what people use.
Built for people and agents. Routed and cached to stay fast and cheap.
Next.jsReactNodeBunGoPythonPostgresRedis
Queues, caching and graceful fallbacks for millions of requests. Deploys, monitoring and patches too: I'm the one who gets paged.
Slack bots, n8n workflows and Apps Script that take repetitive work off a team.
Building the software a real estate investment firm and the organizations it works with run on: investor platforms, learning and operations systems, marketing sites with lead capture, and AI tools. I'm scoping it with the teams, building it, hosting it and keeping it healthy.
Leading the AI engines behind TrueBeep, a marketing platform, and PROA, a multi-tenant ERP: building copilots with memory, retrieval over documents, inbox agents, and model routing and caching that keep cost and latency down.
It started as a Messenger bot that sent lecture notes to classmates. I'm still building it into a full study platform, with a web app, an Android app, analytics, an AI-assisted note pipeline and an MCP server, and I run all of it.
A part-time project I took on alongside my full-time roles. I built most of the frontend for a program management platform: roles and designations, project cohorts and branches, donor and area data, and approvals.
Built WebManza, a drag-and-drop store builder. Moving storefronts to server rendering fixed their SEO, and RTK Query caching cut about 40% of redundant network calls.
My first engineering job. I architected the AdsHobe ads platform frontend from scratch, worked on the DeliveryHobe storefront, and shipped a React Native rider app.
What I do now to keep production AI features cheap and quick: route by difficulty, cache prompts and results, budget tokens, cap spend, benchmark first.
How I built Turbo Reader with a Rust core, SQLite and Tauri 2, kept the installer under 5 MB, and what broke along the way.
How I built MCP servers that let BUTEX students ask Claude or ChatGPT about official notices and NoteBot notes, and why no tool ever throws.
Rust core, React shell, Tauri 2. Under 5 MB to install.
MCP servers for students, with free keys and OAuth.
Shipped the first phase of an internal platform as the sole engineer.
Outgoing connections view and a calmer light theme.
Published the essay and reference design.
AI engine with memory and document search for a multi-tenant ERP.
When I'm not shipping, I'm usually halfway through a TV series, catching a movie or playing a video game, all very casually. And I'll never say no to a game of cricket.
Hiring, or have a product that needs building?
Full-time roles, consulting and advisory work. I usually reply within a day.
afsintripto@gmail.com · @Tripto_Afsin