Trace Smith Email & Phone Number

Houston, Tx, Us

Houston, Tx, Us

Seattle, Wa, Us

Helping energy customers solve complex problems on AWS.- GenAI- ML & Optimization - Data Platform- Networking/Security

Seattle, Wa, Us

• Architected and developed a virtual smart assistant using Amazon Bedrock Agents and LLMs to optimize well planning workflows and accelerating engineers retrieving drilling operations reports/summaries/kpis.• Architected and deployed a real-time drilling advisory system consisting of ETL data pipelines, database management, batch processing, and web application deployment. • Lead data platform developer responsible for implementing a large scale enterprise data platform using a data mesh architecture on AWS. Developed IaC and automation tooling for core platform component; developed automation workflows for data products (producer/consumers). Assisted designing MLOps platform. • Developed a large-scale Predictive Maintenance application to mitigate equipment wear, failures, and unplanned downtime. Deployed real-time anomaly detection pipeline for asset health monitoring; closed loop solution which integrates user feedback for continued model improvement. Tech stack: AWS IoT SiteWise, AWS IoT Events, SageMaker Pipelines, AWS Lambda, and Amazon Managed Grafana.• Developed a scalable, end-to-end, machine learning application to support training thousands of pricing optimization and forecasting time series models. Tech Stack: AWS Batch and SageMaker Pipelines.

Us

• Responsible for designing and productionizing a scalable end-to-end Machine Learning pipeline Azure for high impacting global pricing optimization models (ExxonMobil Chemical Company). • Developed time-series classification model to predict attrition for thousands of customers. Crafted hundreds of lagged features from millions of historical transactions on thousands of customers and key economic indicators to train a ML predictive model. Solution is deployed in Azure.• Implemented infrastructure as code to automate provisioning multiple environments for virtual machines, networking, container registry, and Azure Kubernetes Services (AKS). • Created anomaly detection algorithm for continuous well surveillance monitoring for thousands of oil and gas assets and dynamically routing operators to wells indicating mechanical equipment failures. • Developed a deep-learning time series classification model for predicting downhole drilling non-productive events. Trained models offline, deployed models with REST endpoint, performed real-time inference using Spark Streaming. • Constructed a custom Python algorithm in conjunction with Azure Cognitive Services OCR engine to extract unstructured tables across thousands of scanned PDF documents in order to obtain competitive insights into drilling operations and guide stakeholders with future operations planning.• Developed production-ready code templates to automate deploying Azure Resources and ML pipelines to accelerate delivery times for Data Science Teams. Templates include starter code for ETL and ML tasks with integrated DevOps principles (e.g. CI/CD workflows, unit tests) covering various cloud-based applications: Databricks, Azure ML, Containers, Airflow, and Kubeflow.

Mountain View, California, Us

• Perform code reviews and evaluations of machine learning projects submitted by Udacity students and provide students with Data Science Career consulting and advice; strive to give actionable and helpful feedback to the Data Science community. Projects consisted of Reinforcement Learning & Traditional ML (Regression/Classification).

Houston, Texas, Us

• Research: Geostatistical analysis of microseismic data in fracture modelling; patent pending: https://patents.google.com/patent/GB2569900A/en• Created model using Factor Analysis (FA), a data reduction technique, and Classification and Regression Trees (CART), a predictive data mining technique for predicting oil and gas production of a prospective well; both Factor Analysis and CART, along with geocellular modeling, can then be used to identify optimal well placement and various stimulation and completion characteristics necessary for economic success.Trained CART model to predict if production quality could be related to specific well, seismic, and completion attributes and attempts to explain how input variables relate to a particular phenomenon, like initial or cumulative production.

Houston, Texas, Us

• Research: ‘Pseudo Phase Production Simulation’ – Developed a method to assess multiphase flow production via relative permeability in reservoir flow simulations in order to rank Petrophysical realizations.

Master Of Science (Ms), Data Science

Master Of Science (Ms), Petroleum Engineering

Bachelor Of Science (Bs), Petroleum Engineering