Open to opportunities

Alekhya G

@alekhyag2

Message

Data Scientist building production ML and LLM solutions that drive measurable business impact.

United States

Message

What I'm looking for

I want to build and deploy production ML/LLM systems on AWS with strong MLOps ownership—turning data into measurable business outcomes while collaborating closely with product and executive stakeholders.

I’m a Data Scientist with 4+ years building and deploying production ML models across banking, aviation, and SaaS, working with large-scale datasets on AWS. I’ve delivered meaningful business impact through churn prediction, demand forecasting, anomaly detection, and Generative AI automation, and I consistently partner with executive stakeholders to translate complex models into scalable, revenue-driving decisions.

I’m hands-on with end-to-end MLOps (SageMaker, MLflow, Docker) and LLM-based pipelines, including RAG with LangChain and the OpenAI API. I also bring big data engineering depth (PySpark, Databricks, Hadoop) and I ship production monitoring, dashboards, and CI/CD-backed deployment pipelines so models perform reliably in real operations.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

Santander Bank

Jul 2025 - Present (11 months)

Built churn prediction models on 3.8M customer records, reducing churn from 14.2% to 11.7% and projecting ~$4.5M retention savings. Deployed SageMaker/Glue pipelines and a LangChain + OpenAI RAG classifier, cutting batch inference latency from 4.2 hours to 47 minutes and reducing support triage time by ~50%.

XGBoost Random Forest Churn Modeling AWS Sagemaker AWS Glue MLFlow Pyspark A B Testing Lang Chain RAG

Data Scientist

Frontier Airlines

Mar 2024 - Jun 2025 (1 year 3 months)

Developed ensemble no-show/cancellation models on 1.2M bookings, enabling dynamic overbooking strategies with ~$1.5M projected revenue recovery. Built ARIMA/Prophet forecasting, Isolation Forest anomaly detection, and PySpark on Databricks pipelines, improving MAPE from 9.1% to 4.6% and reducing data availability SLA from 6 hours to 85 minutes.

XGBoost Logistic Regression No Show Prediction Isolation Forest Anomaly Detection Pyspark Databricks SQL Optimization

Data Scientist

MindTickle

May 2021 - Nov 2023 (2 years 6 months)

Built and tuned 7 production ML models for sales readiness scoring, improving accuracy from 71% to 79% with Optuna-based hyperparameter tuning. Delivered recommendation systems and maintained Hadoop/Spark pipelines (99.4% uptime), automating 18 executive reports to cut turnaround from 3 days to 4 hours.