Open to opportunities

István Barra

@istvnbarra

I build machine learning systems, training infrastructure, and evaluation frameworks for telematics and AI products.

At Cambridge Mobile Telematics, I build the core training infrastructure for a foundational telematics model, design model-evaluation frameworks, and optimize training performance to speed model-development iteration.

I lead a team of four data scientists and the company’s generative AI working group. I developed automated labeling for crash-model training data, a claim-adjuster chatbot using open-source large language models, and crash models deployed for four clients.

Previously, I developed quantitative risk forecasting and NLP solutions for thematic investing at BlackRock, and predictive marketing optimization, fraud detection, revenue forecasting, and Bayesian estimation tools at King.

I hold a PhD in Financial Econometrics and have published research on Bayesian analysis, state-space models, high-frequency price changes, and copula models.

Experience

Work history, roles, and key accomplishments

Current

Principal Machine Learning Engineer

Current

Cambridge Mobile Telematics

Feb 2025 - Present (1 year 6 months)

Built core training infrastructure for the foundational telematics model, designed a model evaluation framework, architected codebase refactoring, and optimized training performance.

Python PyTorch AWS Databricks Machine Learning Infrastructure Model Evaluation Performance Optimization

Current

Principal Data Scientist / Team Lead

Current

Cambridge Mobile Telematics

Sep 2022 - Present (3 years 11 months)

Led a team of four data scientists, developed an automated labeling framework for crash-model training data, developed a claim-adjuster chatbot using open-source LLMs, and led the generative AI working group.

Python PyTorch Large Language Models Team Leadership Data Labeling Generative AI

Principal Data Scientist

Cambridge Mobile Telematics

Apr 2021 - Aug 2022 (1 year 4 months)

Developed total-loss, point-of-impact, and crash-severity models, and built a model-training framework supporting rapid development across multiple clients.

Python PyTorch Machine Learning Model Development Crash Modeling

Quantitative Researcher, Vice President

BlackRock

Jul 2018 - Apr 2021 (2 years 9 months)

Developed a new specific-risk forecasting methodology for an equity factor risk model, contributed to a forecast-performance back-testing framework, and built NLP solutions for thematic investing using alternative data.

Quantitative Research Risk Modeling NLP Alternative Data

Principal Data Scientist

King

Mar 2018 - Jul 2018 (4 months)

Developed a predictive machine-learning system for automated marketing-campaign optimization in PySpark and a procedure to detect fraudulent advertising networks.

PySpark Machine Learning Marketing Optimization Fraud Detection

Senior Data Scientist

King

Dec 2015 - Mar 2018 (2 years 3 months)

Built a model with time-varying functional forms for revenue-per-install forecasting, developed a Python package for Bayesian parameter estimation, and evaluated A/B tests using difference-in-differences methods.

Python Bayesian Statistics Time Series A B Testing Difference In Differences

Retail Credit Risk Modeler

K&H Group

Jan 2009 - Aug 2009 (7 months)

Developed and validated probability-of-default and exposure-at-default models, and produced ad hoc management reports on the retail credit-risk portfolio.

Credit Risk Model Validation