Market Analysis & Prediction Engine

I developed a comprehensive options trading analysis system focused on weekly SPY options. The system utilizes Python and TensorFlow to process multiple data streams including market indicators, volume analysis, and technical patterns. The core architecture implements a custom LSTM neural network that processes time-series market data alongside sentiment analysis from financial news sources. The system maintains a rolling database of predictions and actual outcomes to continuously refine its accuracy. A particular challenge I solved was handling the non-linear relationship between option prices and underlying asset movements by implementing a custom loss function that weights predictions based on market volatility. The system integrates with several APIs including Alpha Vantage for market data and NewsAPI for sentiment sources, processing this information through a custom-built pipeline that handles data cleaning, feature engineering, and model training.

Open Source Intelligence Platform

This project emerged from my interest in understanding information flow across social networks. I developed a sophisticated web intelligence platform using Python, Selenium, and advanced NLP techniques. The system can archive and analyze content across multiple platforms, handling JavaScript-rendered content and maintaining session persistence. A key technical achievement was implementing a distributed crawler architecture using Docker containers on AWS Lambda, which allows for dynamic scaling based on workload. The system processes over 100,000 pages daily while respecting rate limits and ethical scraping guidelines. I added computer vision capabilities using TensorFlow to analyze image content and identify patterns across visual media. The backend uses PostgreSQL for structured data storage and Elasticsearch for full-text search capabilities, with Redis handling caching and rate limiting.

Real Estate Valuation System

Building on my interest in both real estate and machine learning, I created an automated valuation system that combines traditional pricing models with modern machine learning techniques. The system ingests multiple data sources including property characteristics, location data, and market trends. I implemented a custom ensemble model that combines XGBoost for numerical features with a BERT-based model for processing textual descriptions. The system achieves 92% accuracy in predicting sale prices within a 5% margin. A particular innovation was the implementation of a computer vision component that analyzes property photos to assess condition and quality factors. The system uses FastAPI for the backend, React for the frontend, and implements a microservices architecture deployed on Kubernetes for scalability.

Autonomous Document Processing Pipeline

This project arose from my work with government documents in the Philippines. I developed a full-stack solution that automatically collects, processes, and analyzes government publications. The system uses a custom OCR pipeline built with Tesseract and OpenCV to handle various document formats, including scanned PDFs and images. I implemented advanced text extraction techniques including layout analysis and table structure recognition. The processed documents are indexed using Elasticsearch with a custom-built search interface that supports both keyword and semantic search capabilities. The system includes a custom-built PDF parsing engine that handles complex layouts and maintains document structure integrity.

Network Analysis Visualization Platform

Inspired by my interest in understanding complex relationships in financial systems, I developed an interactive platform for visualizing and analyzing network relationships among financial entities. The system uses Neo4j as the backend database, with a custom-built API layer in FastAPI. The frontend visualization is implemented using D3.js with custom force-directed graph layouts. A key technical challenge I solved was implementing real-time graph traversal algorithms that can handle networks with millions of edges while maintaining interactive performance. The platform includes custom-built components for community detection and centrality analysis, helping identify key nodes and relationships within complex networks.

Personal Trading Algorithm Framework

As a personal project combining my interests in finance and automation, I developed a systematic trading framework that integrates multiple data sources and trading strategies. The system implements various technical analysis indicators using the TA-Lib library, with custom modifications for specific market conditions. I built a backtesting engine using Python and Pandas that can simulate trading strategies across multiple timeframes and assets. The system includes risk management components that automatically adjust position sizes based on market volatility and account parameters. The framework incorporates a custom event-driven architecture that allows for real-time strategy adjustment based on market conditions.

AI-Powered Social Media Automation Platform

I developed a comprehensive social media management system that leverages generative AI to create, optimize, and schedule content across multiple platforms. The system uses OpenAI’s GPT models through a custom prompt engineering layer that ensures brand voice consistency and content quality. I implemented a sophisticated content pipeline that begins with automated topic discovery using trending analysis and competitor monitoring, followed by AI-assisted content generation and image creation using DALL-E and Stable Diffusion models. The system includes advanced scheduling capabilities that use machine learning to determine optimal posting times based on historical engagement data.

A key innovation was the development of a feedback loop system that analyzes post performance metrics and automatically adjusts the content generation parameters to optimize engagement. The platform integrates with multiple social media APIs and uses a custom-built middleware layer to handle rate limiting and platform-specific formatting requirements. The system successfully increased engagement rates by over 300% while reducing content creation time by 80%. I implemented comprehensive analytics tracking using a combination of PostgreSQL for structured data and TimescaleDB for time-series metrics, with a custom dashboard built in React for real-time monitoring and reporting.

Fraud Detection System

Leveraging my experience in regulatory compliance, I created an automated system for detecting suspicious patterns in financial transactions and business relationships. The system implements graph-based anomaly detection algorithms to identify unusual connection patterns, combined with natural language processing to analyze communication content. I used Apache Kafka for real-time data processing and implemented a custom alert system using Elasticsearch for rapid pattern matching across historical data. The system includes machine learning models trained on historical fraud cases to identify potential risk factors and flag suspicious patterns for investigation.

Open Source Contributions

I actively contribute to several open source projects, focusing particularly on data processing and machine learning tools. Notable contributions include performance optimizations for popular Python data processing libraries and developing new features for web scraping frameworks. My contributions focus on improving code efficiency, adding new functionality, and ensuring comprehensive test coverage. I’ve also created and maintained public repositories containing utilities for web crawling, financial analysis, and data visualization, all built with a focus on scalability and maintainability.