David Ayo
AI/ML Engineer who builds smart apps & automation bots
New to Contra
David is ready for their next project!
Built DROPZONE — a full-stack e-commerce marketplace
with a dark, premium streetwear aesthetic.
Features shipped end-to-end:
-Product catalog with categories, new drops, and
sale sections
-Shopping cart and user authentication system
-Admin dashboard for inventory and order management
-AI-powered product recommendations using Anthropic's Claude API
-Sales analytics with Chart.js visualizations
-Fully responsive across all screen sizes
This project covers the same core architecture needed
for a marketplace - user roles, product/listing
management, search and filtering, and a clean
consumer-facing storefront.
Tech stack: Node.js · Express · JavaScript ·
Chart.js · Claude API · HTML/CSS
0
2
Built an end-to-end AI-driven Forex trading bot that automates the entire trading lifecycle — from real-time market analysis to live trade execution — without any human intervention.
The system connects directly to MetaTrader 5 to pull live OHLCV and tick data, then runs it through a hybrid ensemble AI engine combining XGBoost (pattern detection on tabular indicator data) and LSTM (deep learning for time-series memory). Trades are only executed when both models agree above a confidence threshold — significantly reducing false signals.
Key engineering highlights:
· Smart Money Concepts (SMC) + RSI/ATR indicators feed a complex feature set into the models
· Ensemble predictor with configurable confidence thresholds before any trade fires
· Risk manager calculates position sizes dynamically based on account balance, enforces stop-loss, take-profit, and daily loss limits
· News filter automatically pauses trading during high-impact economic events
· Real-time mission control dashboard (Flask + SocketIO) showing live charts, bot status, open trades, and AI decision logs
· SQLite state persistence — bot resumes exactly where it left off after any restart
This is a production-grade system, not a tutorial project. Every layer — data, analysis, AI, risk, and execution — is decoupled and independently testable.
Tech stack: Python · TensorFlow (LSTM) · XGBoost · Scikit-Learn · MetaTrader 5 · Flask · SocketIO · Pandas · SQLite
1
40
Built and deployed a Hybrid TE-GRU + LightGBM forecasting API for smart grid energy optimization backed by Bayesian MCMC confidence bounds, to ensure physical hardware can react safely to volatile energy demands. Seeing the manual testing dashboard light up and run flawless predictions through the FastAPI backend is an incredible feeling.
1
31
Built a fully automated crypto arbitrage scanner that monitors 5 major exchanges simultaneously every 10 seconds across 20 top coins, firing instant Telegram alerts when a price spread exceeds 1.5%.
The system fetches live prices concurrently from Binance, Coinbase, Kraken, KuCoin, and OKX using async HTTP requests, calculates spreads across every exchange pair in real time, and delivers formatted alerts with exact buy/sell prices and estimated profit per $1,000 invested.
Key engineering highlights:
· Async architecture (Python asyncio + aiohttp) for concurrent exchange polling with zero blocking
· On-chain payment verification - subscribers pay in USDT/USDC and the bot verifies the transaction directly on Polygon/BSC via block explorer APIs, no payment processor needed
· Telegram subscription bot with full admin controls, auto invite link generation, and renewal handling
· Live web dashboard (Flask + Chart.js) showing real-time spreads, alert history, and revenue stats
· Deployed on Railway with persistent storage, graceful shutdown handling, and health check endpoints
Tech stack: Python · asyncio · aiohttp · python-telegram-bot v20 · Flask · SQLite · Railway
1
41
RAG Study Buddy
Built a full-stack, AI-powered study assistant that allows users to upload PDF textbooks and documents, instantly transforming them into interactive learning environments.
The application leverages an advanced Retrieval-Augmented Generation (RAG) pipeline orchestrated with LangChain. Uploaded documents are parsed, chunked, and embedded into a Pinecone vector database. Upon querying, the system retrieves the most semantically relevant sections and passes them to OpenAI's API, ensuring that all answers are strictly grounded in the source text to drastically reduce hallucinations.
Beyond standard Q&A, the system automatically generates study flashcards and maps out document relationships using knowledge graph visualization.
Tech Stack: Python · FastAPI · Next.js · LangChain · Pinecone · OpenAI API
This project demonstrates my ability to build production-ready AI microservices, manage complex backend RAG logic (including semantic chunking and embedding strategies), and integrate them seamlessly with a modern web frontend.
2
56