ABDULMOEIZ
Full-Stack Developer · AI & Automation Expert
What I build with
TypeScript · JavaScript (ES6+) · Python · C++ · C# · SQL · HTML5 · CSS3
Next.js · React · Tailwind CSS · Framer Motion · ReactFlow · Recharts · Figma
Node.js · NestJS · Express.js · FastAPI · REST APIs · WebSockets · JWT · OAuth 2.0 · RBAC
PostgreSQL · Supabase · MongoDB · Redis · Row-Level Security · Schema Design · Indexing
LLM Integration (Claude · Gemini · Groq) · RAG · FAISS · LangChain-pattern agents · Structured JSON · XGBoost · LightGBM · CatBoost · scikit-learn · TensorFlow
Multi-API Orchestration · Playwright · trafilatura · Webhooks · Cron Jobs · Celery · Real-Time Pipelines
Docker · Kubernetes · CI/CD · GitHub Actions · Airflow (Astronomer) · Vercel · Railway
FinTech · Algorithmic Trading · Subscription Billing · SaaS Architecture · GenAI Reasoning Agents
Experience
Designed and shipped arandcolaw.com end-to-end — architecture, schema, backend, frontend, deployment — covering auth, consultation booking, subscriptions, multi-service registrations, case tracking, and a Google-Docs-powered CMS. Serving 500+ monthly active users.
- ▸4-role RBAC on Supabase with Row-Level Security across 20+ tables & 130+ policies — zero unauthorized-access incidents in 5+ months.
- ▸Lemon Squeezy recurring + one-time billing with HMAC-SHA256 webhook verification and idempotent event routing; 200+ monthly transactions reconciling at 99.9%.
- ▸Google-Docs CMS with auto-generated SEO (slug, meta, OG + JSON-LD) and 60s ISR for blog posts and case studies.
- ▸p95 API latency 720ms → 210ms via composite indexing, RLS optimization, and query tuning; Lighthouse 62 → 94.
Built a full-stack social media marketing app centralizing 5+ accounts per user, with content scheduling, automated posting, and unified analytics.
- ▸Node/Express services with OAuth across Meta Graph (Facebook + Instagram), LinkedIn Marketing, and Twitter APIs — token refresh + account linking.
- ▸MongoDB models supporting ~10k scheduled posts in test.
- ▸Reusable React dashboard components cut duplicate UI ~40% and new-feature rollout from 3 days to 1.
Selected work
Entropable — AI Crypto Trading Platform
A production full-stack SaaS — a visual strategy builder, ML market-event signals, walk-forward backtesting and portfolio-grade risk controls in one workflow. Built over a layered Express backend (Controller → Service → Repository → DB) with 20+ domain services and Python ML spawned via child process or Celery. No Python, no quant team required of the user.
- ▸CryptoMind AI — RAG reasoning agent — FAISS top-3 retrieval over a curated setup store + live Binance candles; Redis-cached follow-ups run ~15× faster than a naive re-run (~45s → ~1–3s).
- ▸Visual Strategy Builder — ReactFlow no-code canvas; the LLM edits strategy graphs through structured JSON patch operations, with in-canvas backtest preview before deployment.
- ▸ML Pipeline — 132 indicators → 8 calibrated event classifiers → an AUC-weighted XGBoost + LightGBM + CatBoost soft-vote ensemble, with t-1/t-2/t-3 lag features.
- ▸Backtesting & Validation — Walk-forward with Combinatorially Purged Cross-Validation (purge + embargo); Sharpe / Sortino / max-drawdown; DEAP genetic optimizer with risk-constraint penalties.
Pre-Call Brief Generator
An end-to-end Python automation pipeline that turns a prospect URL into a 1–2 page pre-call intelligence brief, replacing 20–40 minutes of manual research per lead with an automated Markdown + PDF deliverable emailed to the salesperson.
- ▸Multi-API collection — 7+ free-tier sources — Playwright site/LinkedIn scrapers, Meta Ad Library, PageSpeed, regex tech-stack detection, DIY SEO signals — with graceful degradation.
- ▸6-stage LLM workflow — Gemini 2.5 Flash/Pro with Pydantic-typed structured output: parallel asyncio collection → extraction → marketing-posture analysis → brief synthesis, with automatic data-gap flagging.
- ▸Async + delivery — Celery + Redis jobs, FastAPI ingestion with API-key auth + webhook callbacks, SQLAlchemy/Alembic schema for step-by-step observability, WeasyPrint PDF + Resend email.
- ▸Swap-ready — Production-shaped under ~$25 total — every free-tier component is a single-file swap to a paid alternative without touching pipeline logic.
Automated DEX Trading Bot
An automated TypeScript trading bot for decentralized exchanges executing momentum and volatility strategies with dynamic take-profit, trailing stop-loss, and adaptive risk management.
- ▸Real-time ingestion — DexScreener pipeline (axios + ethers.js) consuming on-chain price, volume, and liquidity across 100+ token pairs, filtering for volume spikes, liquidity changes, and new listings.
- ▸Trade logging — Every entry, exit, indicator value, decision, and PnL persisted row-by-row to structured .xlsx via Danfo.js + ExcelJS for post-run analysis.
- ▸Anti-FOMO logic — Pullback + breakout confirmation reduced false entries ~35% vs. naive momentum (measured across simulated runs); reacts to price spikes in <500ms.
Med Spa Call Agent — "Maya"
An AI-powered voice receptionist for U.S. medical spas — a real-time speech → LLM → speech pipeline with mid-call tool execution and a live operator dashboard streaming call transcripts as they happen.
- ▸Real-time voice pipeline — Low-latency speech → LLM → speech loop that holds a natural conversation with callers in real time.
- ▸Mid-call tool execution — Function/tool calling lets Maya book appointments, answer FAQs, and capture leads while the call is still live.
- ▸Live dashboard — A Next.js operator dashboard streams live call transcripts over Server-Sent Events (SSE) for monitoring and handoff.
Side projects & experiments
LET'S BUILDTOGETHER.
Open to full-stack and AI engineering roles — remote or in Islamabad.