Machine Learning Engineer

Engineered and deployed gradient-boosted and deep-learning fraud-detection features into Radar's real-time scoring pipeline, supporting risk decisions for millions of daily transactions at sub-100 ms latency. Built more than 40 behavioral and network features for a transformer-based payments foundation model, improving fraud recall by 2.4% for a high-risk merchant segment while maintaining the false-positive rate.

Designed an LLM-powered agent that translates plain-English fraud policies into validated executable rule code, reducing rule-authoring time from 30 minutes to under 3 minutes across more than 200 rules. Optimized Adaptive Acceptance authorization-rate models, recovered 1.1% of legitimate payments, and deployed a chargeback-prediction model that improved dispute win rates by 9%.

Established offline and online evaluation, drift monitoring, and A/B testing processes with risk and data-engineering partners. Presented business impact to senior stakeholders while supporting PCI-DSS and SCA/3DS compliance requirements.

Completed a Machine Learning Engineer internship supporting payments, fraud, and risk machine learning initiatives at Stripe.

Contributed to the transition into a contract Machine Learning Engineer role, applying machine learning expertise to real-time risk decisioning and production model development.

Productionized a clinical NLP entity-extraction pipeline using Vertex AI and the Healthcare Natural Language API, improving extraction F1 from 0.74 to 0.88 and reducing manual chart-review effort by 40%. Developed OCR and Document AI solutions for medical and invoice documents, processing more than 1 million pages monthly at 96% field-level accuracy.

Trained deep-learning demand-forecasting models for a retail/CPG client, reducing MAPE from 18% to 11% across more than 500 SKUs. Fine-tuned an LLM-based clinical summarization prototype on Vertex AI, reducing clinician documentation time by 35% in a pilot across three hospital units.

Owned end-to-end MLOps, including Docker-based training and serving, Kubeflow and Vertex Pipelines, drift detection, and automated retraining that reduced model-refresh time by 60%. Led client workshops, translated requirements into solution designs, and applied HIPAA and Responsible AI bias checks.

Prototyped CNN and transformer baselines for image- and text-classification proofs of concept. Benchmarked six pretrained computer vision and NLP architectures to evaluate accuracy-latency trade-offs for client engagements.

Built interactive Python and Looker Studio dashboards for model performance and data-quality visibility. Automated data-ingestion and feature-engineering pipelines with Python and BigQuery, reducing dataset-preparation time by 50% and earning promotion to Machine Learning Engineer.

Master of Science in Computer Science.