§ drishti
Creator, sole builder & operator
Apr 2026 – Present
Claude Opus 4.8
Reasoning-first tool-calling
Python
systemd on VM
Cloudflare R2
Next.js / Vercel
Creator · Live AI Trading Signals
Drishti
Live LLM signals · 15-min cycle · 8 markets · real capital.
Win rate (closed set)
Perp markets
Decision cycle
PageSpeed (landing)
Tab. 1 — key metrics
§ 01
The idea
Drishti asks a frontier LLM one disciplined question every cycle: given everything the market is showing right now, is there an edge — and if so, where are entry, stop and target? The model reasons first, then emits a structured decision the system can act on.
It runs a 15-minute, regime-aware cycle across 8 Delta crypto perpetual markets, with 30+ engineered features per asset (CVD, IV term structure, basis, global long/short, OI by exchange, Coinbase premium and more).
§ 02
From signal to real orders
A live executor turns accepted signals into real, risk-managed orders on Delta India — reduce-only stop-loss and take-profit brackets, fixed sizing, a guardian process, and hard kill-switches.
The publisher runs as a systemd service on a VM, writes snapshots to Cloudflare R2, and a Vercel front-end renders signals, the equity curve and closed-trade stats continuously.
§ 03
What it demonstrates
Prompt and context engineering under real money, cost-aware model routing, and an operational loop that has to be right every 15 minutes — not just in a notebook.
Fig. 2
Drishti live equity curve — balance, drawdown, CAGR, Sharpe with win/loss stats
Fig. 3
Drishti live signal feed — direction, entry, stop, target per market
Fig. 4
How Drishti works — the 15-minute decision-to-execution cycle
§ colophon
Stack
Claude Opus 4.8 · Reasoning-first tool-calling · Python · systemd on VM · Cloudflare R2 · Next.js / Vercel · Delta Exchange API