Sameed Shah
AI Engineer building agents, apps & automations.
A mobile finance tracker that rolls salary, side income, a MEXC crypto balance and money lent or borrowed into one honest net-worth figure. Highlights include a purpose-built amount keypad, recurring income rules that post themselves, read-only MEXC sync with a history chart and loans that close themselves at zero. It also has a monthly budget tagged Essential / Planned / Optional, so it's obvious what to cut first. There's a privacy mode, a biometric lock, light and dark themes, and JSON export. It's built with Expo (React Native) + Supabase.
0
36
I designed and built Cloudex Technologies as an AI-native company. It's staffed by 18 "Digital FTEs": AI agents with written roles, real tools and a reporting line. They sit in four departments: the Office of the CEO, Delivery, Inbound & Demand, and Outbound & Sales. They run on Paperclip with Claude Code as the main lane and Codex as an adversarial reviewer. The whole company is defined in a git repo (constitution, governance, per-agent instructions), so the runtime is a deploy, not a pet.
The rule it's built on: author ≠ verifier ≠ releaser. No agent approves its own work, every build passes five quality gates, and a human principal releases every outbound batch, which can be done by replying SEND or HOLD to an email digest. The sales department runs its own AutoGTM engine. It sources leads from the free Companies House register and researches each prospect's website, keeping only facts quoted word for word. It then drafts short, human emails and sends them within warm-up limits, with auto opt-outs and bounce stops. It runs privately on a VPS over a Tailscale mesh, with budget caps and a full audit trail.
1
1
56
The website for Cloudex Technologies, where I build "Digital FTEs": AI employees that take repetitive work off a team. I designed and built it with Next.js and Sanity CMS, so the team can publish insights and case studies without touching code. The service pages cover workflow automation, voice agents, marketing automation and AI data analysis, and Cal.com (http://Cal.com) booking is embedded. The isometric "AI office" hero and soft motion keep a technical offer friendly and easy to follow.
0
25
Basheyes is the control room for an AI trading-signal engine I built and run on a VPS. Every hour it scans the Binance futures market. Five analysis agents score each coin: technical, open interest/volume, order book, supply/demand and sentiment. A risk agent sizes the trade, an orchestrator writes the reasoning, and qualifying signals go straight to WhatsApp.
The dashboard shows it all live. A radar plots every coin by score and bias, outcome gauges track win rate and realized R, a heat matrix breaks down each agent's vote, and the signal feed follows every trade to TP, SL or expiry. It's Next.js + Tailwind v4 with lightweight-charts, streamed over WebSocket from the engine's own API. The public copy runs on Vercel behind a password gate and a Cloudflare Tunnel, so the trading API is never exposed to the internet. So far it has logged 650+ signals and 750+ WhatsApp deliveries.
0
34