Fatih Hekimoğlu

Aiming for a 2:1. Selected final-year marks: Software Hut 69, Human-Computer Interaction 67, Functional Programming 66, Speech Processing 64, Software Development for Mobile Devices 63.

Leakage-safe walk-forward ML pipeline over UK equities; hyperparameters committed to a SHA-256 manifest before evaluation. Dependence-aware multiple testing (Clark-West, Romano-Wolf, Model Confidence Set) over an iid/HAC/ Politis-White variance envelope; O(T²) prefix-sum BSADF. Honest finding: rejections collapsed from 14/74 to 2/74 across variance estimators and the long-short signal failed the deflated-Sharpe test; reported as an audited predictability claim, not a trading strategy. Python, scikit-learn.

AI/ML news-analysis platform for a free and unfiltered press ("same news, different worlds"); ~2,200 monthly site visits as of June 2026, early stage. Real-time pipeline over 144 Turkish outlets: three-method ML story-clustering ensemble (TF-IDF similarity, 4-gram fingerprints, entity overlap), news sentiment analysis, who-said-what views across bias zones, blindspot and cross-spectrum surprise detection. Next.js 16, React 19, TypeScript, Supabase, pgmq, Deno 2, Vercel cron; event-driven, no long-running workers. tayf-ruby.vercel.app · github.com/noxire-dev/tayf

Full-time summer placement. Built the firm's public website from scratch (PHP, JavaScript, HTML, CSS) and a CMS admin panel so non-technical staff publish without developer involvement. Server and power-backup systems; documented and tested a disaster-recovery plan; modernised company devices.

Led a 6-person team building a client book-reading web platform (SQL, JavaScript, Ruby/Sinatra); Engineering Without Borders transport feasibility study (Cambodia).

Rebuilt from first principles and judged out of sample; ~1,200 tests across the research code. Includes GARCH-vs-ML volatility, HMM market regimes, and options pricing NN vs Black-Scholes.

From-scratch mean-variance optimization. 562 tests · 99% coverage · reproduces He-Litterman to 1e-4.

Cointegration stat-arb. 527 tests · OOS deflated Sharpe 0.71 vs IS 1.94; the out-of-sample number is the headline.

Rigorous evaluation harness; honest null. 82 tests · 97% coverage · Jensen's α p=0.69; cannot reject zero skill.

PCA + Marchenko-Pastur RMT signal/noise separation (MP edges to 1e-12) · Fama-French 3/5/6 + Carhart with Newey-West HAC by default · end-to-end algo-trading system (Alpaca, Dockerised) · RAG agent over SEC EDGAR 10-K filings.

Team judged by the client to have delivered the most effective software under an agile process; group lead for an AI-assisted Kanban board built for the consultancy Reply (COM3420, mark 69).

BetterPicked: CNN computer vision to cut supermarket food waste ~40%; led financial modelling, market research and the pitch. Also received the Dr Trish Murray Professional Behaviours Team Award.