Projects
Q_Trader – Modular Execution Framework for Binance
- Developed a lightweight trading framework focusing on the execution layer (order placement, balance management, market data retrieval)
- Designed a plug-in interface for strategies via a factory pattern, allowing seamless switch between demo and proprietary strategies
- Live trading loop supports both Binance Testnet and mainnet, with history logging and periodic strategy re-initialization
- Proprietary strategies include adaptive weight models trained with PyTorch, which remain private
- Released under MIT License (executor open-sourced, proprietary strategies kept private)
Quommoty – Futures Pricer (QuantLib-based)
A prototype futures pricing framework built in C++ with QuantLib. Implements a shared pricing interface and specialized pricers for crude oil (cost-of-carry model) and electricity (forward-anchor with delivery averaging).
This is an iterative prototype project, primarily for personal learning and practice.
- Focus: commodity futures pricing models (storable vs. non-storable commodities)
- Includes demo tests and modular design for extension
HBitGuard – On-chain Event Monitoring Toolkit (BSC)
A lightweight Python toolkit for monitoring and decoding on-chain events on BNB Smart Chain (BSC). It is designed for tracking transaction flows to analyze abnormal signals or potential Web3 attacking events.
- Supports minimal ABIs for common standards: ERC20, ERC721, Uniswap V2/V3, WBNB
- Built-in event registry with handlers
- Transaction tracker that decodes logs into normalized events and records unknown raw events
- Real-time runner loop with block windowing, safe head confirmations, and transaction hash deduplication
- Extensible architecture: custom ABIs and event handlers can be easily integrated
Survival Modeling with DeepSurv (QRT Challenge)
- Modeled patient survival using linear CoxPH (sksurv) model and a PyTorch-based nonlinear DeepSurv model
- Focused on censored data handling, model interpretability and calibration
Image processing - Obstacle and Road Marking Detection via Classical Stereo Vision
This project implements a pipeline for obstacle and lane marking detection from stereo camera input, using traditional computer vision methods such as disparity estimation, edge detection, and Hough transforms.
Ray Tracer
Realistic rendering with reflection and shadows using recursive ray tracing.
View full project
Petit Metro Simulator
Simulates train behavior under CBTC-based signaling systems, modeling tracks, signals, and station control.
Includes extensible train dynamic modules, train control system logic, and visualization of simulation results.
Below is a sample output showing a train adapting to speed limits and signal targets along the track.
View full project
GitHub repo