Senior Data Scientist · Risk Analytics

Prateek
Parida

Turning complex financial data into credit decisions, fraud defences, and compliance intelligence. 3+ years delivering ML solutions across consumer lending at scale.

Credit RiskFraud DetectionNLPA/B TestingPython · SQLML Pipelines
0%
Portfolio Loss Reduction
0%
Compliance Review Time Saved
0M+
Transactions Modelled
0+
Years in Risk Analytics

Risk & ML,
end to end

I'm a Senior Data Scientist specialising in credit & fraud risk analytics with hands-on experience at Discover Financial Services. I build ML pipelines that go all the way — from raw behavioural and transactional data to production-deployed models informing policy for millions of consumer accounts.

My work sits at the intersection of statistical rigour and business impact: designing A/B tests that justify strategy changes, building NLP tooling that scales compliance review, and migrating fraud signals to cloud-native infrastructure in real time.

I hold an M.S. in Information Systems from the University of Utah (David Eccles School of Business) and a B.A. in Finance from Portland State University.

Risk & Analytics
Credit UnderwritingLoss Rate OptimisationPortfolio MonitoringFraud StrategyA/B TestingBayesian Methods
Machine Learning
XGBoostLogistic RegressionIsolation ForestAutoencodersNLP / LLMARIMARAG
Platforms & Tools
PythonSQLSnowflakeAWS SageMakerDatabricksAirflowdbtTableau

End-to-End
Risk Intelligence

Three production-grade projects spanning the full spectrum of financial risk: credit underwriting, compliance NLP, and real-time fraud detection.

01 · Credit Risk

End-to-End Credit Risk Modelling Pipeline

Built a full-lifecycle probability-of-default (PD) model on large-scale behavioural and transactional datasets across consumer card and personal loan accounts. Owned every stage — data ingestion, feature engineering, A/B evaluation, and production deployment on AWS SageMaker. Risk cohort targeting cut default rates by 10–15% and reduced portfolio loss exposure by 25%.

XGBoostLogistic RegressionPythonSQLSageMakerSnowflakeAirflowA/B Testing
Feature importance — PD model · hover for values
25%
Portfolio Loss Reduction
Measured against pre-model baseline across card & personal loan books
15%
Default Rate Drop
In targeted high-risk cohorts via evidence-backed A/B policy changes
0.81
Model AUC-ROC
Production champion model validated on hold-out consumer segments
E2E
Full ML Lifecycle
Feature engineering → training → evaluation → SageMaker deployment
02 · NLP · Compliance

NLP Complaint & Compliance Monitoring Engine

Automated analysis of unstructured fraud complaints and UDAAP/Metro-2 escalation records at scale, replacing manual triage workflows. Used transformer-based NLP pipelines to classify complaint categories, detect emerging client experience gaps, and surface compliance signals — cutting manual review time by 50%.

spaCyTransformersBERTPythonRegex PipelinesSnowflakeUDAAPMetro-2
Complaint classification confidence · click bars to explore
50%
Review Time Reduction
Manual compliance & complaint triage cut in half via NLP automation
10K+
Records Monthly
Unstructured escalation records analysed at scale
93%
Classification Accuracy
Validated against labelled compliance ground truth set
Faster Gap Detection
Client experience gaps surfaced twice as quickly
03 · Fraud Detection

Real-Time Fraud Detection: ATO & CNP Models

Designed and deployed ML-based anomaly detection for Account Takeover (ATO) and Card-Not-Present (CNP) fraud, migrated to Snowflake for real-time inference. Used Isolation Forest and Autoencoder architectures modelling normal transactional behaviour across velocity, device fingerprint, and geo-behavioural signals.

Isolation ForestAutoencodersPythonSnowflakeReal-Time InferenceDatabricksPattern Recognition
Precision / Recall tradeoff · hover to explore threshold
95%
Fraud Recall Rate
ATO & CNP fraud caught by ensemble anomaly detection
RT
Real-Time Scoring
Model outputs embedded into Snowflake for live monitoring
2
Fraud Vectors Covered
ATO and CNP fraud in a unified pipeline
↓FP
False Positive Reduction
Geo-behavioural features reduced false alerts vs. rule-based baseline

Career
Timeline

From intern to senior data scientist, every role focused on measurable risk impact.

Discover Financial ServicesJan 2023 – Dec 2025 · Chicago, IL
Senior Data Scientist — Credit & Fraud Risk
  • End-to-end credit underwriting models — PD modelling, feature engineering, production deployment — reducing portfolio loss exposure by 25%.
  • A/B testing frameworks evaluating strategy changes; delivered policy recommendations cutting default rates 10–15% in targeted cohorts.
  • NLP scripts automated compliance review of UDAAP/Metro-2 escalation data, saving 50% of manual triage time.
  • Migrated fraud detection strategies to Snowflake; embedded real-time anomaly detection outputs for live monitoring.
  • Tableau dashboards tracking portfolio KPIs and credit exposure, presented to senior leadership.
Energy & Geoscience Institute, U of UtahJul 2022 – Dec 2022 · Salt Lake City, UT
Data Scientist
  • Time-series forecasting and anomaly detection on oil & gas production datasets, improving issue detection speed by 35%.
  • Influenced drilling and recompletion decisions for projects valued at $50M+, reducing costly production downtime.
LendioJan 2022 – Jul 2022 · Lehi, UT
Data Scientist Intern
  • Analysed 9M+ financial transactions to build borrower segmentation models, improving loan targeting precision by 25%.
  • Cluster outputs integrated into loan origination workflows, reducing early-payment-default risk on new SMB loans by 10%.

Let's
Connect

Open to senior data science and risk analytics roles. Based in the US. Feel free to reach out.