class EmmanuelAdutwum:
def __init__(self):
self.role = "Software Engineer · ML Engineer · Quantitative Researcher"
self.education = {
"university": "Soka University of America",
"major": "Economics & Mathematics",
"mit_micromasters": "Statistics & Data Science (MITx)"
}
self.experience = ["Wells Fargo", "CNO Financial Group", "CSIR Ghana"]
self.software_engineering = [
"Distributed real-time systems (WebSocket, event-driven architecture)",
"Full-stack development (React, FastAPI, REST APIs, PostgreSQL)",
"Systems design (scalability, low-latency, data pipelines)",
"Object-oriented & functional (Python, C++, Java, TypeScript)",
"Cloud & DevOps (AWS, Azure, Docker, CI/CD)",
]
self.machine_learning = [
"Deep Learning (LSTM, Transformers, Attention, YOLOv5, PointNet)",
"Classical ML (SVR, Random Forest, Gradient Boosting, Clustering)",
"NLP & Vision (sequence models, 3D point cloud processing)",
"MLOps (model serving, experiment tracking, deployment)",
]
self.quantitative_finance = [
"Signal processing (Kalman Filter, OU Process, ADF stationarity test)",
"Volatility models (GARCH, Hurst Exponent, EWMA)",
"Options pricing (Black-Scholes, Greeks, Delta Hedging)",
"HFT / Market Making (Avellaneda-Stoikov, limit order book dynamics)",
"Position sizing (Kelly Criterion, CVaR, Monte Carlo simulation)",
]
self.passions = ["Deep Learning Research", "Systems Engineering",
"Quantitative Finance", "Open-Source ML"]
def say_hello(self):
return "Hi — I engineer production systems, train models, and build at the intersection of code and mathematics."
me = EmmanuelAdutwum()
print(me.say_hello())A production distributed system: event-driven data pipeline → real-time WebSocket API → React SPA with a quantitative signal engine running live on MetaTrader 5
|
|
System Design:
MT5 COM Bridge (Windows)
│ 1s polling loop
▼
FastAPI (async, uvicorn)
├── /ws ──── WebSocket broadcast ──── React SPA (useWebSocket hook)
├── /api/ohlcv/{symbol} CandlestickChart + Kalman overlay
├── /api/analytics/monthly P&L calendar heatmap
├── /api/analytics/session 7×24 session heatmap
├── /api/analytics/correlation Pearson matrix (Pandas pivot)
└── /api/analytics/rolling Rolling Sharpe & Sortino
│
Telegram Bot API
(alert state machine, 5-min cooldown)
|
$25,000 Award Competitive grant for cryptographic security research in FinTech — applied to blockchain-based authentication protocols |
Published Mathematician Two accepted problem solutions in Pi Mu Epsilon Journal — USA's premier undergraduate mathematics journal (#1131, #1385) |
MIT MicroMasters Statistics & Data Science — graduate-level program covering probability theory, inference, ML, and data analysis |
Production Trading System M3 Ultra Alpha runs live on MT5 across 5 instruments — Kalman, GARCH, Hurst, Kelly all operating in real markets |
|
ML Benchmark Result 0.723 R² on California Housing via optimised SVR — 4-stage ablation study, leakage-safe pipeline, +0.123 gain over published baseline. Submitted to Springer Nature. |
Industry Experience Financial services & research across Wells Fargo, CNO Financial Group, and CSIR Ghana — real-world quantitative & engineering work |
HFT Market Making Engine Built Avellaneda-Stoikov market maker in C++ with live limit order book, inventory risk management, and spread optimisation |
Real-World Finance Tool Black-Scholes delta hedging simulator for Ghana Cocoa Board — live demo with volatility surface, Greeks, and AI price forecasting |
| Area | Demonstrated Through |
|---|---|
| Distributed Systems | M3 dashboard: WebSocket pub/sub, async FastAPI, event-driven data pipeline with real-time state management |
| API Design | RESTful endpoints with FastAPI (OHLCV, analytics, health); clean separation of transport and business logic |
| Frontend Engineering | React 18 SPA with custom hooks (useWebSocket), TradingView chart integration, component-driven architecture |
| Data Pipelines | MT5 → Python bridge → Pandas analytics → JSON → UI; handles OHLCV, trade history, account state |
| Algorithms & Math | Published Pi Mu Epsilon journal solver; Kalman filter, Hurst, Kelly Criterion from first principles in Python |
| Low-Level Systems | HFT market making engine in C++; limit order book, Avellaneda-Stoikov spread optimisation |
| ML Engineering | SVR (0.723 R², submitted to Springer Nature), LSTM, Transformers, YOLOv5/PointNet for Tesla sensor fusion — model training to deployment |
| Cross-language | Production code in Python · C++ · MQL5 · JavaScript/React · R · Java — pick up any stack quickly |
|
C++17 · Zero ML libs · Adam · BatchNorm · Dropout · ~97% MNIST
|
Multi-head causal attention · Weight tying · Top-k/p sampling · Cosine LR
|
||
|
92%+ CIFAR-10 · ONNX Export · C++ Runtime · <1ms inference
|
Lift-Splat-Shoot · CenterPoint · KITTI 3D mAP · Mixed Precision
|
||
|
31+ dB PSNR · Mip-NeRF IPE · Marching Cubes · 360° Video
|
Black-Scholes · Avellaneda-Stoikov HFT · Kelly · GARCH · Hurst · OU · Cocoa Derivatives
|
Black-Scholes · Delta Hedging · Real-time Greeks · Commodity Finance
|
Avellaneda-Stoikov · Limit Order Book · Inventory Risk · Market Microstructure
|
|
SVR · Feature Engineering · Ablation Study · 0.723 R²
|
Cryptography · Blockchain · Financial Security · $25,000 Grant
|
||
|
VaR/CVaR · Cornish-Fisher · Monte Carlo 10K · Sharpe/Sortino/Calmar · Macro Regimes
|
RFM Segmentation · Cohort Retention · CLV · Churn Prediction AUC 0.84 · K-Means
|
| Publication | Venue | Link |
|---|---|---|
| Optimised Support Vector Regression for California Housing Price Prediction: The Critical Role of Feature Engineering and Hyperparameter Tuning | Under Revision · Targeting Discover AI · Springer Nature · 2026 | |
| Equilateral Triangle Geometry Problem | Pi Mu Epsilon Journal #1131 | |
| Continued Fractions, α-Fibonacci Numbers, and the Middle b-Noise | Pi Mu Epsilon Journal #1385 |