ML / AI Engineer · London, UK

Priyansh

I build machine-learning systems for finance and fraud.

Junior ML engineer building finance ML end to end — a credit-card fraud detector and a leakage-controlled equity ranker. Open to ML/AI roles in finance.

See selected workAbout

Selected work

What I'm building

Machine-learning systems across finance and computer vision, documented end to end — the problem, the approach, the tradeoffs, and honest results, including where they fall short.

RankAlpha

Live

Built the ML, then beat it. A cross-sectional equity ranker that learns to order the S&P 500 — then a pre-registered duel showed plain 12-1 momentum wins the long-only product book (after-cost Sharpe 1.79 vs 1.41), so momentum ships and the ML stays research. The model's ordering skill is real, but it only looks monetisable long/short. Built on a SEC point-in-time fundamentals gate, a 380+ test suite, and published negative results.

  • Python
  • LightGBM · LambdaMART
  • Learning-to-rank
  • Walk-forward + embargo
  • Pre-registered duel
  • SEC EDGAR point-in-time
  • SHAP
After-cost Sharpe — momentum vs ML
1.79 vs 1.41
ML mean Rank IC (vs momentum)
0.0505 vs 0.0414 (t=1.64)
Audit machinery
380+ tests · SEC point-in-time
Live productResearch demo (Streamlit)

FraudLens

Live

Explainable credit-card fraud detection that prices model errors in £, finds where the model breaks, and proves the ceiling is the data — not the algorithm.

  • Python
  • XGBoost
  • SHAP
  • Imbalanced data
  • Streamlit
Live productResearch demo (Streamlit)

StockAI Terminal

Research

An earlier exploration: an ML + LLM prototype that predicts a single stock's direction — AdaBoost signals validated on a chronological split with a 30-day embargo, explained by an LLM fed structured model output so it cites rather than hallucinates, and never gives buy/sell advice. The direction-prediction approach is the one I later measured as a dead end and superseded with RankAlpha's cross-sectional ranking; it is kept here because that failure is what forced the better design.

  • AdaBoost
  • Chronological split + embargo
  • Llama 3.3 · Groq
  • GitHub Actions
  • Streamlit
Live demo

Anti-UAV Detection

Research

A cascaded drone-detection system (detector → bird-verifier → type-classifier) and a co-authored paper — submitted to BCS SGAI 2026, under review — that measures, honestly, what survives domain shift: near-perfect in-distribution scores that partly collapse on unseen data, behind a perceptual-hash leakage audit.

  • Computer vision
  • YOLO11
  • ResNet-18
  • EfficientNet-B0
  • Cross-domain eval
  • BCS SGAI 2026 · under review

About

I'm Priyansh, a First-Class Computer Science graduate and junior ML engineer based in London, UK. I'm focused on machine learning for finance — fraud, risk, and the unglamorous data work that makes a model trustworthy in production.

Lately I've been building two systems end to end: a credit-card fraud detector that prices its errors in pounds, and RankAlpha, a cross-sectional equity ranker with a leakage-controlled, cost-aware backtest. I care more about honest metrics and clean pipelines than leaderboard scores.

I'm looking for an ML/AI role where I can learn from senior engineers and ship real systems — ideally at a UK fintech or bank.

Contact

Get in touch

I'm open to ML/AI roles and happy to talk about the work. The fastest way to reach me is email.

priyansh2005p@gmail.comGitHubLinkedInRésumé