I build ML models that help banks catch fraud as it happens.

Data scientist working on real-time fraud detection and risk models for banks in the UAE and Oman.

Get in touch See my work

About

What I actually work on.

I'm Aditya, a data scientist with an engineering background. I work at the intersection of machine learning and financial risk. For the past two years I've built real-time fraud detection systems for banks across the UAE and Oman, models that run in production and score hundreds of thousands of transactions a day.

My work covers the whole pipeline: pulling apart raw data, engineering behavioral features, cutting dimensionality, validating models, and getting them live. I care about how ML gets used in financial decisions more broadly, not just fraud but credit risk and lending too. I like problems where a model's output actually changes what a business does next.

Experience

Where I've worked.

Effiya Technologies

Ahmedabad, India · – Present

  • Ahmedabad, India
  • Abu Dhabi, UAE
  • Muscat, Oman

  1. – Present

    Senior Analyst

    Ahmedabad · On-site in Muscat, Oman

    • Developed an ML-based payment fraud detection model for a major UAE bank, achieving ~80% Gini on test data and ~75% Gini out-of-time, now scoring ~300K transactions daily in production.
    • Built the end-to-end model pipeline in Python and SAS: data quality checks, feature engineering across 60+ behavioral variables, dimensionality reduction, and iterative training.
    • Implemented Python Signature classes for real-time behavioral profiling and deployed the model package to production.
    • Conducted an on-site Fraud Resilience Assessment in Muscat, analyzing fraud risk exposure across retail and corporate banking to strengthen monitoring and rule frameworks.

  2. –

    Analyst

    Ahmedabad · On-site in Abu Dhabi, UAE

    • Designed and implemented real-time fraud detection logic, including customer behavioral profile algorithms across multiple banking channels, processing 500K+ transactions daily across all channels at up to 150 TPS.
    • Collaborated on-site with cross-functional teams and business stakeholders in Abu Dhabi, aligning analytical logic with evolving business requirements.
    • Designed automated analytical workflows for performance measurement, KPI reporting, and rule evaluation at scale, which helped cut false positives by 40%.

  3. –

    Analyst Intern

    Ahmedabad

    • Analyzed the Dynamic Customer Risk Assessment (DCRA) model to evaluate drivers of customer risk scoring and identify enhancement opportunities.
    • Wrote SAS scripts to extract model-impact metrics, generating insights that supported fine-tuning of customer risk thresholds in production.

Education

B.Tech, Electronics & Communication

Indian Institute of Information Technology Surat

– · CGPA 9.20

Selected work

Things I've built.

Python · SAS · Predictive Modelling

Real-time payments fraud detection

A logistic regression model that scores every transaction on a 0–999 risk scale for a UAE-based bank. It holds ~75% Gini on out-of-time data, using 10 features shortlisted from 60+.

  • Ran end-to-end data quality checks on customer and transaction data: fill rates, outlier capping, and distribution checks.
  • Engineered behavioral features spanning velocity, ticket size, time-based aggregates, and beneficiary patterns.
  • Applied PCA-based dimensionality reduction and functional transformations to handle multicollinearity.
  • Developed Python signature classes for real-time serialize/deserialize and update logic on behavioral profiles.
All projects on GitHub (opens in a new tab)

Toolkit

What I work with.

Languages

  • Python
  • SQL
  • SAS
  • C++

Modelling

  • Predictive modelling
  • Logistic regression
  • Classification
  • Scorecard development
  • Feature engineering
  • Dimensionality reduction (PCA)

Validation

  • Model validation
  • Gini / KS statistics
  • Out-of-time testing
  • Threshold tuning
  • Data quality checks

Fraud & risk

  • Real-time transaction scoring
  • Behavioural profiling
  • Rule tuning
  • False-positive reduction
  • Customer risk assessment
  • KPI reporting

Libraries & tools

  • numpy
  • pandas
  • scikit-learn
  • matplotlib
  • seaborn
  • pickle
  • Jupyter
  • Git

Domain

  • BFSI
  • Payments
  • Fraud detection
  • Risk analytics
  • Credit risk & lending

Let's talk.

Working on an interesting problem in financial modelling, risk or analytics? I'd like to hear about it.

jadhavaditya517@gmail.com