What I build
Production systems that run unattended. Not prototypes — things that handle real money, real data, real decisions.
Business processes that run themselves. Product data conversion, email management, lease auditing, deal alerts — built once, runs forever.
n8n workflows / Python / cron / API integrations
Multi-agent systems that think and act. 11 AI agents running 24/7 across Discord, WhatsApp, and custom interfaces — with memory, learning, and LLM routing.
Claude / GPT / Qwen / custom orchestration
Scraping, synthesis, and intelligence. Real-time social monitoring, API data pipelines, automated research reports across 10+ sources.
REST APIs / web scraping / LLM synthesis
End-to-end algo trading. Signal generation, Kelly criterion sizing, circuit breakers, multi-server execution. Live on Hyperliquid with real capital.
Python / 3 VPS nodes / 2,000+ live trades
My AI assistant
I built an AI system I use every day. Here's what it looks like in action.
Real screen recording. WhatsApp + Kaleo AI.
Text or voice message my AI from anywhere. It responds in seconds with full context of my systems.
Reads and sends emails via Gmail API. Manages two accounts. Categorizes, prioritizes, and cleans inbox automatically.
Persistent memory across conversations. Knows my projects, preferences, trading rules, contacts, and history.
Writes in my exact tone — emails, documents, messages. Five voice tiers from casual to formal. Passes as human.
Deploys code, restarts services, monitors health, and runs trading operations across three VPS nodes. 24/7.
Not one chatbot — a team. Research, trading, email, monitoring, content, and operations agents running concurrently.
Control everything from your phone. Slash commands, voice messages, file sharing. Works from anywhere.
Live proof
My trading system runs on real capital. This equity curve pulls live from the Hyperliquid exchange API.
Indexed to 100. Real capital. Real fills. No backtests in disguise.
Under the hood
After 10 years of manual trading without consistent results, I built a multi-strategy algo trading engine. Every safety feature exists because something broke without it.
Maker orders on Hyperliquid at 0.5bps. ATR-based trailing stops activate at +0.5% profit and trail 2% from peak. Three independent layers of stop-loss protection on every position.
Fractional Kelly (25% of raw) caps sizing at 1-5%. Circuit breaker trips when consecutive losses exceed binomial probability. 5% daily loss = global stop.
Below 4h = noise. 49% WR makes money if TP:SL math works. Taker fees kill edge at small capital. Win rate does not equal profitability.
Let's build something
Whether it's automating a business process, building an AI agent, or setting up a trading system — I ship production code that runs without you.