FinnewsHunter: Multi-agent financial intelligence platform powered by AgenticX. Real-time news analysis, sentiment fusion, and alpha factor mining.

Quantitative investment strategy capturing divergence between analyst ratings and news sentiment using FinBERT

ML-driven alpha signal generation for U.S. equities. Ridge, Random Forest, and XGBoost trained with walk-forward validation. XGBoost achieves Sharpe 1.07 out-of-sample. Full feature importance and IC stability analysis.

VLLM-based trading system

GPU-Accelerated NDI Broadcast Engine with Desktop Capture, NDI Main + Backup Input, Main Backup Failover and NDI + Decklink Out

Implemented and adapted selected alphas from the WorldQuant 101 Alphas framework to the cryptocurrency market. Built a 2020–2025 dataset with train/test split, evaluated signals using IC statistics, combined weak alphas via rank normalization, and backtested portfolio strategies, achieving improved Sharpe over an equal-weight benchmark.

v3min8e

A repository containing 8 alpha signals with a signal combination engine incorporate a risk model, with 76+ automated tests.

Developed a comprehensive alpha generation and portfolio optimization system combining traditional factor models with alternative data sources, achieving consistent risk-adjusted outperformance over market benchmarks through sophisticated signal blending and dynamic risk management.

Momentum trading strategy for crypto using ML (XGBoost) vs traditional logic — with backtesting, VaR risk analysis & Monte Carlo simulation.

Zero-infra alerter for Congress trades + SEC 13D/G + Asian 5%-stake disclosures. Runs on GitHub Actions cron, DMs Telegram.

An Intraday Alpha Backtesting Framework for popular equities