CTO in Web3 and DeFi. Board member. 15+ years in engineering, 7+ of them leading teams.
I run engineering for a Web3 company and hold a formal board seat. My focus is the intersection of three things that rarely sit together in one operator: production-grade EVM and DeFi systems, classical backend craft at scale, and AI-native engineering workflows that are reshaping how high-performing teams actually ship.
Lead a 5 to 15 person engineering organization across backend, smart contracts, and frontend. Own architecture, delivery, hiring bar, security posture, and the operating rhythm of the team. Set the technical direction for products that move real value on-chain, where mistakes are public, expensive, and permanent.
Board responsibilities cover technology strategy, security oversight, and engineering risk at the governance level.
Most engineering leaders sit on one side of a clear line: pure Web2, pure Web3, or pure AI. I operate across all three, and my last few years have been about turning that overlap into a competitive advantage.
Web3 teams that adopt AI-native workflows responsibly will outbuild the ones that do not, and the gap is widening every quarter. Most of the industry has not figured out yet what responsible looks like in a domain where a bad merge can drain a treasury. That is the problem I spend most of my attention on.
7+ years in CTO and Head of Engineering roles. Shipped EVM smart contract systems into production. Built and rebuilt backend platforms on Node.js, TypeScript, Postgres, Redis, and the standard AWS and GCP stack. Led teams through full delivery cycles: greenfield, scale-up, incident response, security review, audits, and the slow grinding work of paying down technical debt without breaking what already runs.
Operating principle is consistent across all of it: correctness first, maintainability second, speed earned through both. Production-grade on the first pass.
Agentic development workflows in serious engineering organizations. Specifically, how teams that ship financial infrastructure can use AI agents as real contributors without losing the rigor that the domain demands.
I am building this in two places at once. Inside the company I lead, where the bar is production and the stakes are real. And on three personal products that double as a sandbox for the same workflows under conditions I fully control.
Transcribr: macOS menu bar app that captures mic and full system audio into a single file and pipes it through Whisper. SwiftUI, ScreenCaptureKit, real-time audio mixing at 48 kHz.
GpsLogger: end-to-end GPS tracking system with a SwiftUI iOS client and a typed backend. Design rule: collect raw location data with zero interpretation. No trip detection, no behavior inference, no opinions about the data.
ValenciaGo: multilingual event discovery platform for Valencia. Aggregates from six public sources, normalizes, deduplicates, classifies, summarizes via AI in English, Ukrainian, and Spanish, delivers through a Telegram bot.
Each one is small on purpose. The point is to test where AI-assisted workflows hold up outside a controlled demo, with a real user (me) and a real maintenance burden.
- GitHub: @athlonUA
- Email: alexgarmatenko@gmail.com