Experience
2025 — Now
2025 — Now
United States
Partner with product and analytics teams to translate customer engagement challenges into measurable ML objectives, defining success
metrics that improved campaign targeting precision by 46% and increased conversion lift by 52% across enterprise accounts.
• Architected scalable data pipelines using PySpark and Airflow to process 12M+ behavioral events daily, reducing preprocessing latency
by 58% and improving feature reliability consistency by 49% across model training environments.
• Designed and trained transformer-based recommendation models, implementing structured feature selection and evaluation workflows
that increased personalization accuracy by 44% and reduced model drift occurrence by 41% over quarterly monitoring cycles.
• Deployed containerized inference services through Docker and Kubernetes on AWS SageMaker, lowering average API response time by
47% and improving deployment stability metrics by 63% during peak traffic periods.
• Established experiment tracking and model version governance using MLflow, strengthening reproducibility controls by 54% and
decreasing debugging turnaround time by 48% during production performance investigations.
• Implemented real-time monitoring dashboards tracking drift, latency, and cost utilization, enabling proactive optimization strategies
that reduced cloud inference expenses by 42% while maintaining performance benchmarks within defined SLA thresholds.
2021 — 2023
2021 — 2023
India
• Collaborated with platform engineering teams to convert workflow automation requirements into predictive modeling initiatives,
increasing incident resolution accuracy by 51% and decreasing ticket routing errors by 43% across enterprise service modules.
• Developed structured ETL frameworks integrating transactional logs and service metadata, improving dataset integrity validation by 57%
and reducing manual preprocessing effort by 62% across recurring training pipelines.
• Built gradient boosting and neural network classification systems for ticket prioritization, raising precision-recall balance by 48% and
shortening average handling time by 45% across high-volume support categories.
• Introduced controlled hyperparameter experimentation processes with cross-validation governance, improving model stability variance
by 44% and minimizing retraining frequency by 40% across quarterly release cycles.
• Delivered REST-based inference endpoints integrated into internal service applications, increasing automation throughput by 59% and
decreasing manual assignment dependency by 53% across regional support operations.
• Designed performance monitoring frameworks capturing prediction quality, latency distribution, and system utilization metrics,
improving operational transparency by 61% and accelerating root-cause identification by 47% during anomaly investigations.
Education
New Jersey Institute of Technology