Automation Engineer focused on AI-driven workflows, API integrations, and autonomous agents. I've designed, built, deployed and monitored 10+ real systems solo — most of them live and clickable below. Learning fast, shipping faster.
From scattered manual processes to reliable, monitored pipelines. I handle the full stack — integration, logic, deployment, and ongoing maintenance.
n8n, Make.com, Zapier, and custom Python pipelines. CRM automation, lead gen flows, invoicing, onboarding, notifications — the boring stuff that costs businesses hours every week.
Autonomous agents built on Claude & OpenAI. Multi-model pipelines with fallback, RAG, tool use, and human-in-the-loop. Telegram/Discord bots that actually do work — monitoring, trading, customer support, data collection.
REST/GraphQL APIs, webhooks, OAuth, cron jobs, Cloudflare Workers, D1/KV/R2. Real-time data sync, ETL pipelines, web scraping, custom dashboards. If two systems need to talk, I wire them up.
Live systems doing real work right now — not demos. Green dot = health check passing. Private systems show as case studies — full demos on request.
An autonomous AI agent city. A fleet of LLM agents earn standing, run an economy, and build real websites & games inside the world — multiplayer, persistent, on the edge.
No-code AI agent builder — design an agent visually, train it on your docs, then deploy to Telegram, a web widget, a hosted page, or a REST API. D1 + KV, managed credits.
A 24/7 autonomous AI agent that monitors markets, executes trades across Solana / pump.fun / Polymarket, and reports via Telegram. Multi-model with fallback, real-money pipelines, self-healing infrastructure. The ultimate automation case study.
LLM-driven Android control over ADB — an agent that operates a real phone: reads the screen, taps, types, completes tasks. AI agency on physical hardware.
A team of LLM agents — founder, scout, builder, warden and more — that claim jobs from a shared board, talk to each other, and ship real work. A dispatcher routes each task to the right agent and model over a typed event bus, with turn-taking rules.
Cron-driven operations for a whole fleet of bots — scheduled agent missions, hourly usage and health reports pushed to Telegram, and automated digests, all running unattended on a Raspberry Pi.
A supervisor process running several real Telegram bots — each with its own persona and LLM — handling owner-only DMs plus group @mentions, with crash-proof pollers kept alive under PM2.
A live "stock market" for the agent fleet — every agent is priced like a coin, with candlestick charts generated from its real activity. Prices, trades and a running ledger update automatically every 15 minutes by cron.
Internal ops dashboards for the agent fleet — a shared "brain" board and a live status bar — that rebuild themselves from real activity on a cron, so the view is always current with zero manual updates.
A provisioning site that gives every agent a real on-chain Solana wallet — keys generated and vaulted server-side, with any fund-moving action guarded behind explicit confirmation and dry-run by default.
A live-data product tracking the GTA VI creator economy — real Steam & market stats, served from a single Worker with its own tiny API and cron jobs.
An offline-first companion PWA for the New Forest — installable, works with no signal, full app polish. Demonstrates real PWA craft end to end.
Automated ops dashboard monitoring 15+ PM2 services. Health checks, auto-restart on failure, deployment tracking, log aggregation, and alerting — all running hands-off on a Raspberry Pi.
Real-time paddleboarding conditions — live tide & weather APIs, web-push alerts, and a daily cron forecast, all from one same-origin Worker.
A power-user reference PWA for shipping real software from a Chromebook — AI, agents, the Claude SDK, CLI, SSH and deploy, all in one place.
Product & landing page for a mobile cockpit that drives Claude Code from your phone — design, copy and marketing polish, shipped on its own domain.
On a single Raspberry Pi I built a fleet of LLM agents, each with its own persona, model and job. They claim tasks from a shared board, talk to each other, monitor the systems, scout opportunities and ship real work — and I supervise the whole thing from my phone over Telegram.
Sets direction, breaks the week into jobs, and hands each one to the right agent.
Hunts passive-income and market signals, then ranks them into a running ledger.
Turns approved ideas into shipped code, pages and tools — end to end.
Runs health checks and money-path guardrails; catches failures before I do.
Packages progress into updates and posts them to the team and the channel.
Keeps the shared board and logs current so nothing gets lost between sessions.
Throws in the lateral, creative takes the others would never reach for.
Today they run my studio — research, builds, monitoring and reporting, 24/7, with barely a hand on the wheel. Tomorrow, this is the automation I'd stand up inside your team: agents that watch your systems, make the small decisions, and ship — so your people spend their time on the work only people can do.
It started as a way to give myself a team of one. It's become proof I can design, deploy and run autonomous systems that actually earn their keep.
Edge-first. Everything below is in something I've shipped — not a wishlist.
Every project above is real and deployed — most are live and clickable right now. I learn by building things that run in production, not by collecting tutorials.
From API integration to database to frontend dashboard, I've taken projects the whole way solo — so I understand how the pieces fit and can pick things up quickly.
LLM orchestration, tool use, multi-model fallback, RAG — I've built real autonomous agents. It's where I'm strongest and where I most want to grow.