MercuriusLite
A Lightweight Quantitative Investment System
MercuriusLite is a self-built quantitative system focused on asset allocation of macro asset classes in personal investment portfolios. It provides objective guidance for strategy developers through advanced analytical tools and systematic approaches.
Macro Asset Allocation
Systematic approach to allocating capital across major asset classes including equities, bonds, commodities, and alternatives.
Portfolio Optimization
Quantitative methods for optimizing risk-adjusted returns based on individual investment objectives and constraints.
Strategy Development
Framework and tools for developing, testing, and implementing systematic investment strategies.
Risk Management
Comprehensive risk assessment and monitoring capabilities for portfolio management.
Tracking Automation
Automated systems for continuous monitoring of portfolio performance and market signals.
Message Delivery
Real-time notifications and reporting delivery to keep investors informed of critical updates.
Oculus
Objective Market State Probability Assessment System
Oculus is a probability statistical model based on multi-source, massive data, designed to assess market status. Upholding a worldview of simulability rather than predictability, Oculus employs statistical indicators, AI methods, and Monte Carlo simulations to objectively assess market states.
Probabilistic Modeling
Utilizes advanced statistical indicators and Monte Carlo simulations to model market probabilities.
AI Integration
Leverages artificial intelligence methods to process multi-source massive data for deep insights.
Objective Assessment
Provides unbiased market state evaluations based on rigorous quantitative analysis.
Wealth Position
Two-way wealth percentile explorer
Convert net wealth into an estimated percentile—or a percentile into a wealth threshold—for the World, Mainland China, United States, and Hong Kong. Explore yearly estimates, statistical units, model intervals, and data-quality caveats in one interactive view.
Explore Wealth Position →FIRE Calculator
Monte Carlo Retirement-Ruin Simulator
An interactive Financial-Independence / Retire-Early calculator. Tune your expected annual return, strategy volatility, inflation, and withdrawal rate, and watch the probability of running out of money evolve over a 5–50 year horizon — computed live in your browser via Monte Carlo simulation.
Monte Carlo Engine
2,000 lognormal (geometric) return paths simulate portfolio survival under the fixed-real "4% rule" withdrawal strategy.
Interactive Inputs
Adjust annual return, volatility, inflation, and withdrawal rate with live sliders — the ruin curves recompute instantly.
Coverage Sensitivity
Ten curves show how external cash flow covering 0%–100% of spending reshapes your probability of ruin.
MCL Doctor
Data Pipeline Health Monitor
MCL Doctor is an internal observability tool that checks the freshness of every data file in the pipeline — both at the OS level (file modification time) and at the data level (last Date record in the CSV). Results are grouped by category and served as a live JSON dashboard.
Two-Layer Freshness
Each target passes two independent gates: file mtime and last CSV record date, each checked against a per-frequency latency window.
YAML Registry
All 130+ monitored files are declared in a single monitor.yaml with group, frequency, latency override, and date column per item.
Pattern Matching
Supports {yyyymmdd} filename-embedded dates, * wildcards, and ** recursive glob for option chain and yield curve files.
cmip6-to-wrfinterm ↗
CMIP6 to WRF Regional Downscaling Pipeline
An automated one-stop pipeline for converting CMIP6 global climate model output into WRF-compatible intermediate files, enabling regional dynamical downscaling research. Handles multi-layer soil data, non-standard calendars, and supports multiple GCMs.
Automated Conversion
End-to-end pipeline from CMIP6 NetCDF output to WRF intermediate format with minimal manual intervention.
Multi-GCM Support
Compatible with MPI-ESM, EC-Earth3, CESM2, and other major CMIP6 global climate models.
Complex Data Handling
Robust handling of multi-layer soil data, non-standard calendars, and varying grid configurations.
pyCFRAM ↗
Climate Feedback-Response Analysis Method
A Python implementation of the Climate Feedback-Response Analysis Method (CFRAM), decomposing climate model temperature changes into individual physical drivers. Designed for CMIP6-scale attribution studies with dual radiation engines and multiprocessing acceleration.
Physical Decomposition
Decomposes temperature changes into contributions from CO₂, water vapor, clouds, aerosols, surface albedo, and other drivers.
Dual Radiation Engines
Supports both RRTMG and Fu radiation schemes, with multiprocessing acceleration for CMIP6-scale datasets.
Attribution Analysis
Rigorous quantitative attribution of climate model outputs to individual physical forcing and feedback mechanisms.