• Realized $5M annually in LPO by designing and scaling CV pipelines for drone image analysis, including EDA, training, inference, and promotion pipelines (single and multi-task), and deploying ML models to production across 3 business units.• Increased Chevron’s competency in data science, statistics, and ML by mentoring 16 engineers (5 years).• Enabled enterprise capabilities in work order completion (RAG), emission prediction (regression), KPI extraction, corrosion detection (TensorFlow), curve analysis (time-series), and pricing decision analysis.• Built automated data pipelines for extracting, transforming, and loading (ETL) structured and unstructured big data across applications and Azure Data Lake (Apache Spark and SQL).• Reduced image classification inference time by 90% via migrating model deployment (TensorFlow) and inference API from Databricks Jobs to serverless Azure Functions.• Led productionization of alarm flood management system, optimizing code and ML pipelines, reducing alarm floods by 70% across 3 process areas, and driving rapid experimentation from PoC to production.

Performed exploratorily data analysis and cleaning (Python, Pandas) on a large public transportation dataset, and built a graph (NetworkX) with time-series congestion signals defined on the nodes to be used for testing big data (spatio-temporal) predictive models• Investigated data quality issues, modeled time-series signals and node relations from the raw data, and supported decisions by developing advanced visualizations, plots, and time evolution maps, and investigating spatial correlations• Improving query time, memory-limitation, and availability by migrating to the Hadoop ecosystem (PySpark) and AWS

• Designed and developed a production deployed thread-safe Go package for streamlining the instrumentation and exposing Go micro-services’ performance metrics using the Prometheus framework via HTTP/2• Ensured maximum maintainability, readability, and reusability by developing unit test coverage (100%), and including extremely detailed comments and robust sets of examples using the agile framework (scrum)• Proposed and developed a machine learning solution (Python, supervised) for reducing downtime, and validated it by building server/client (Go) micro-services (shell scripting/docker compose for automating building docker images and running containers)

• Acted as a machine learning specialist in four machine learning projects for fault detection and optimization• Collaborated in a cross-functional and remote team for testing, validation, and embedding predictive models on edge• Developed the experiment plan, specified motion profiles, motors and other equipment, and built the setup• Programmed the PLC, conducted the tests, gathered raw signals, and performed data pre-processing• Built supervised learning models based on various algorithms in Python for classifying coupling conditions• Improved accuracy by feature extraction, PCA, filtering, and composed a Jupyter Notebook to report the results