Backend engineer. Cybersecurity systems by day, AI-agent-built products on the side.
I work on Python services, messaging and infrastructure inside complex operational systems, lead frontend architecture for our control interfaces, and spend my evenings shipping small products with a crew of AI coding agents.
Side projects, built mostly evenings and weekends, many of them with AI agents doing a share of the implementation. Open one to see more.
Guests scan a QR code or open a link and shoot straight from their phone's browser, with optional filters. No app to download and no guest account.
Photos are collected and delivered privately to the couple after the event. No live feed, no social layer, just the feeling of a disposable camera on every table.
The goal: type a short idea and get back a 3–6 minute Hebrew episode, shaped as a natural conversation between two hosts.
The app is a static Next.js export, so generation runs in a separate worker that handles the long-running jobs and credentials. Episodes publish in the app first, with a human approving the early ones.
A personal assistant agent running on OpenClaw on a VPS, managed with PM2, talking to me through a Telegram bot and able to use local models through Ollama.
Next to it I run several coding agents in parallel, including Claude Code and Hermes. That workflow led to research on a cross-vendor dashboard for monitoring concurrent coding agents.
Learners practice reading real chart patterns in short, game-like rounds instead of watching another course.
The market research pointed away from consumers and toward a white-label product that trading instructors can offer their own students.
A Hidden Markov Model that segments price history into hidden regimes and compares how those regimes behave across different assets.
Day-to-day tools at work and on side projects.
Python backend services in cybersecurity and operational systems, cross-component troubleshooting, and architecture of the React and TypeScript control interfaces. Mentoring the developers around me.
C/C++ and Python for firmware work, with Python tools that automated validation.
Machine learning, deep learning, optimization and NLP.
Operational role in the Intelligence Corps during regular service; multiple operational deployments in reserves.