Hi, I'm Rastra — a Data Engineer with a long-standing interest in the crypto markets. This project started as a way to put both sides of that together.

Two years ago I asked a simple question: when is a genuinely good time to open a position? What followed was anything but simple.

My first instinct was to trade frequently — more signals, more upside. It didn't work. Win rates hovered around 55%, and transaction fees eroded every gain. I was optimising for activity rather than conviction. So I changed approach: fewer trades, bigger moves. Swing trading — waiting for the pendulum to build real momentum before acting.

That conviction needed evidence. I backtested the model across two years of Bitcoin data, tuning the statistical rules until I had something defensible:

Not perfect — crypto is inherently uncertain — but a strong signal in a chaotic market.

There was still one gap: the bot couldn't see what I could see. Sometimes a trade was almost there — close to triggering, but not quite. So I built in human override, letting me act on that judgement when it mattered. That's a big part of why the win rate you see today is where it is.

What started as something personal made me think: what if someone without a technical background could benefit from the same signals? All they'd need is a Telegram notification — buy here, sell here. Before I open that up, I want solid, real-world evidence. A live bot, a simulated portfolio, transparent results. That's what you're looking at right now.

Along the way I've applied everything I've picked up as a data engineer — Kafka streams, Azure Key Vault, XGBoost, Supabase, FastAPI, and Claude AI to interpret each trade. It's become a genuine engineering project. And proof that if you're curious enough, very little is actually out of reach.