TL;DR
• Micro- to milli-second latency targets (C/C++, perfect hash functions, inotify)
• Quant models (PDE solvers, neural nets)
• Language-learning tools (personal, performance-obsessed)
📈 models – Quant Finance Research Suite
(Open-source core repo is public; private components marked.)
🔗 github.com/sg-hk/models
▶ ML Options (Python)
- 3-layer dense NN (12–15 neurons each)
- Trained on GBM Monte-Carlo and analytic BSM PDE outputs
▶ PDE Solver (C++)
- Crank–Nicolson finite difference scheme with Rannacher smoothing
- Vectorised grid solver for stability and precision
▶ Crypto Arb (C++) (private)
Details (click to expand)
- Spot-futures arbitrage engine across crypto exchanges
- Profitable and benchmark-beating; currently live
- Implements inventory-aware quoting & latency-hedged fills
▶ Kelly & Macro Dashboards (Python) (private)
Details (click to expand)
- Reproduction of “Empirical Asset Pricing via Machine Learning” (Kelly et al., 2018)
- Extended for maximum return strategies
- Built macro dashboard tools (Clewell et al., 2018) for trade idea generation & backtesting
barbar – Ultra-Low-Latency Status Bar (C)
🔗 github.com/sg-hk/barbar
- Desktop status bar built with no external dependencies
- Reacts instantly to file events via
inotify, avoiding polling
- Shared memory + single mutex = sub-millisecond latency
- Mirrors constraints of real-time market data UIs
ccq – Chinese SRS & Dictionary CLI (C + Go)
🔗 github.com/sg-hk/ccq
| Module |
Stack |
Highlights |
| Reviewer |
C |
Custom FSRS model, 0 dependencies, fully built from scratch |
| Lookup TUI |
Go |
Perfect hash on >900k entries, Bubble Tea TUI, instant lookups |
Why it matters to trading: Hash-table design, memory-mapped I/O, ultra-responsive CLI/TUI interfaces.
✉️ Contact
If you’d like a walkthrough or access to private code under NDA: