Hello, I'm

Steve Duval

Software Engineer & ML Engineer

Principal Software Engineer and tech lead with 9 years of experience building reliable, high-performance systems. Over 7 years guiding architectural design, observability, and system performance across agile teams. Passionate about bridging the gap between machine learning research and production — currently pursuing an M.S. in Computer Science (ML) at Georgia Tech with a 4.0 GPA.

Get in touchResumeLinkedInGitHub

Experience

Principal Software Engineer

Liberty Mutual Insurance

Jun 2017 — Present
  • Progressed from Software Engineer to Principal Engineer, demonstrating consistent technical growth and expanding scope of leadership.
  • Architected and developed Java-based pricing engine APIs that dynamically rate customer insurance policies using telematics and driving behavior data.
  • Built a data pipeline to ingest ML models and utilize them in production, leveraging Apache Spark.
  • Leading and mentoring a team of 9 engineers across architecture reviews, design decisions, and code quality.
  • Built and maintained a Python ETL data processing pipeline that ingests database changes, validates and cleanses data, then calls downstream APIs.
  • Built fully asynchronous, low-latency APIs for external and internal users using Spring Microservices, WebFlux, AWS Lambda, and Kafka.

Business Systems Analyst

Liberty Mutual Insurance

Jun 2013 — Mar 2017
  • Identified and analyzed trends from structured production database data and unstructured customer comment data.
  • Became the subject matter expert for data querying on the eService team using SQL.
  • Participated in gathering and processing data from A/B tests and customer surveys.

Education

M.S. Computer Science — AI/ML

Georgia Institute of Technology

2025 – Present

GPA: 4.0

Artificial IntelligenceMachine Learning for TradingComputing for Data AnalysisDeterministic OptimizationData Analytics in Business

B.S. Finance

University of New Hampshire — Whittemore School of Business

2009 – 2013
StatisticsDerivative SecuritiesQuantitative Decision MakingInvestment AnalysisMacro/Micro EconomicsFinite Mathematics

Skills

Languages & Frameworks

JavaPythonCSQLSpring / WebFluxPandas

ML / AI

RegressionKNNEnsemble BoostingReinforcement LearningNeural NetworksTime-Series ModelingML Pipelines

Data & Systems

Apache SparkETL PipelinesDistributed SystemsDatabase SystemsKafka

Cloud & DevOps

AWS (EC2, Lambda, DocumentDB)CI/CDObservabilityInfrastructure as Code

Research

Effects of LLM Use on Critical Thinking and High-Level Cognitive Functions

Unpublished2026

An IEEE-formatted systematic literature review examining the long-term effects of LLM usage on critical thinking and its underlying cognitive components — executive functions and metacognition — and how these effects vary across different LLM usage types.

IEEELiterature ReviewLLMCritical ThinkingCognitive ScienceAI & Society

Carrot or Stick: Reinforcement Learning in Investing

Unpublished2025

Designed and backtested a Reinforcement Learning agent for automated investment decision-making. Explored reward-shaping strategies across equity and forex time-series data, benchmarking the RL agent against regression, KNN, and ensemble boosting models. Unable to publish paper and source code publicly, this was a graded course project (CS 7646).

Reinforcement LearningAlgorithmic TradingMachine LearningFinance

Projects

Paper Implementation and NBA Prediction Model

2025

Re-implemented the model selection framework from Walsh & Joshi (2024), comparing accuracy vs. calibration as selection criteria for identifying edge in NBA game outcome prediction markets on Kalshi. Trained and evaluated multiple ML classifiers, finding that well-calibrated probability estimates — rather than raw accuracy — better identify profitable opportunities in live contracts.

Machine LearningSports BettingPrediction MarketsPython
Walsh & Joshi (2024)

Contact

Interested in collaborating or have a question? Feel free to reach out.

[email protected]
LinkedInGitHub