Mohamed Rodani

Mohamed Rodani

Mohamed Rodani

Physics undergraduate at Queen Mary University of London with a particular interest in machine learning research, quantitative systems and scientific computing. Open to 2026–27 internships and spring weeks in ML engineering or quantitative research, in London.

Selected work GitHub Request CV

Selected work

Flagship · Quantitative ML · Team of two · Mar–Apr 2026

QMML Market Making Hackathon

Nine live rounds: predict a hidden fair value from an anonymised feature table, quote a bid and ask, then trade against whichever team became market maker. The edge came from quoting wide, sizing up only where cross-validation backed a strong linear fit, and trading the minimum everywhere else.

simulated bankroll, from £100,000
£305,727
behind 1st place
£1,153
rounds traded at minimum size
6 of 9

Mine: prediction models, model selection, strategy and the tiered sizing logic. Kieran Cooke: EDA, feature engineering and the Kelly implementation.

Read the case study → Code

Model selected per stock and the sizing tier it justified
StockModelR² (in-sample)Sizing
1Linear regression0.98Big
2Linear regression0.96Big
3Linear regression0.93Moderate
4Gradient boosting0.27Minimum
5Ridge (α = 100)0.20Minimum
6Mean baseline0.00Minimum
7Ridge (α = 100)0.04Minimum
8Ridge (α = 100)0.04Minimum
9Mean baseline0.00Minimum
Nine synthetic stocks, one round each, as selected by strategy.py. Sizing: big ≈ 15% of bankroll at risk over a 2·RMSE move, moderate ≈ 5%, minimum = 10 shares. Cross-validated RMSE per stock is in the case study.

ML systems · Team of two · December 2025 → 2026

Gweizy — gas-price forecasting for the Base network

Winner of the QMUL AI Society × Coinbase hackathon: a live pipeline that samples Base gas prices every five minutes and serves 1 h / 4 h / 24 h forecasts from a Flask API to a React front end. Only the one-hour forecast beats chance, and the case study says so.

1 h directional accuracy, against a 50% baseline
59.8%
R² at 1 h: a weak but real signal
0.071
five-minute samples by mid-December 2025
26,128

Mine: backend, data pipeline, models, API and deployment. Senan: React front end.

Read the case study → Code

  1. Base RPCbase fee per block, rotating endpoints
  2. Collector5-minute sampling → PostgreSQL
  3. Featureslags, rolling stats, momentum, cyclical time
  4. Models1 h / 4 h / 24 h regressors + spike classifiers
  5. Flask APIREST + Socket.IO, TTL cache
  6. Front endReact / TypeScript on Cloudflare Pages
Data path from chain to decision interface. Backend on Railway; models retrained from the collected history with a holdout baseline check.

NLP · Solo project · 2025

BERT financial sentiment — optimising for the class that matters

Fine-tuned bert-base-uncased on 5,842 financial sentences, 15% of them negative; the first model scored 81.5% accuracy while missing 71% of negative examples. Class-weighted loss largely fixed that for two points of accuracy, at a cost in negative precision.

negative recall, up from 28.6%
87.4%
negative precision, the cost
52.8%
labelled sentences, 15% negative
5,842

Read the case study → Code

Validation set (1,169 sentences), before and after class weighting
MetricOriginalImproved
Negative recall28.6%87.4%
Negative F10.410.66
Accuracy81.5%79.4%
Single 80/20 split, seed 42. Accuracy went down; the metric that matters for a bad-news detector went up.

Tabular ML · Solo entry · November 2025

Kaggle Playground S5E11 — loan-default prediction

LightGBM tuned with Optuna and scored on the same five stratified folds throughout. The write-up keeps the ideas that failed next to the one that helped.

private leaderboard AUC
0.92450
final rank, 950th of 3,850
Top 25%
ideas that made CV worse
2 of 3

Read the case study → Code

Out-of-fold ROC-AUC, five stratified folds
ExperimentCV AUCDecision
Tuned LightGBM baseline0.923351Keep
+ engineered ratio features0.922625Drop
LightGBM / XGBoost / CatBoost blends0.922904 /
0.922999		Drop
5-seed average + external statistics0.923682Submit
The submitted run scored 0.92450 on the private leaderboard, about 0.0008 above its CV estimate.

Background

BSc Physics, Queen Mary University of London (2025–2029, expected). A-levels in Mathematics, Further Mathematics and Physics.

Contact

The fastest route is email. I reply to every message about roles, research or collaboration.