Barret Miller

Experienced technical leader in the emerging technology, innovation, and analytics fields with a track record of excellence in execution and a bias toward action.

• Directed a team of 40 developers including 3 engineering managers across AI, DevOps, and Software Engineering practices.
• Deployed GenAI based semantic vector search to TysonFoodservice.com drastically improving search performance in the first external facing application of GenAI for Tyson Foods.
• Deployed GenAI Agentic assistant solution AskDEB to all 130,000 Tyson team members in April 2024. AskDEB was a multi-model, multi-cloud integrated microservice architecture with multiple tools and integrations to internal systems and knowledge repositories.
• Deployed multiple GenAI Pilots and POCs across Tyson business functions: AskFill (NL to SQL for Fill Rate), AskLIMS (NL to SQL for FSQA pathogen testing), SPHINX (NL to expense data anomalies), OneTyson Comment Analytics (survey sentiment analysis effect on turnover and absenteeism), SCRIBE (contract redlining for legal), PRISM-AI (contract grading for procurement), LUCID (answer questions on USDA export requyirements)
• Lead a team of 8-15 developers and analysts working within a time boxed structured innovation process that we worked as a team to define across dozens of proof of concept and pilot emerging technology projects.
• Deployed Tyson Food's first production computer vision application for product recognition into production at 8 poultry plants. Utilized Tensorflow deep learning framework along with SageMaker, Lambda, and Step Function AWS services to automate a machine learning pipeline for the labeling of data and training of a 98% accurate image classification model, Deployed application as a set of containerized microservices on Kubernetes in the cloud and at the edge/fog in our plants.
• Lead development of computer vision application utilizing ArUco tags and OpenCV to track racks of inventory in cold storage warehouse.
• Lead development of computer vision applications utilizing MXNet GluonCV deep learning framework for detecting defects in poultry shackles, detecting defects in breaded and fried product (85% IoU), counting trays of inventory in racks (95% IoU), and packages of product on conveyor belts (98% IoU), and counting chicks and chickens in hatcheries (96% IoU).
• Lead development of a computer vision application to detect breaks in plastic conveyor belts utilizing realtime 3D line scan data. Compared PointNet neural net for direct pointcloud inference with traditional approach of flattening 3D data into 2D heatmaps for inference. Solution utilized an LMI Gocator 3D scanner.
• Lead development of a robotic solution for data collection of cooked product used to train a deep learning model to predict core temperature from infrared surface temperature readings. Solution utilized Cognex 3D line scanner, a FLIR thermal camera, AWS GreenGrassv2 for edge container and application management.
• Lead development of computer vision applications for mask detection and social distance monitoring at a poultry plant utilizing Palantir Foundry and Artificial Intelligence Platform (AIP).
• Lead Development of TySim, a home grown solution for generating synthetic data for training of computer vision applications. Solution made use of docker container running Blender 3D headless and scripting the randomization of mesh geometry as well as PBR textures on 3D chicken nugget models to dynamically generate thousands of images of scenes of product on conveyor labeled for supervised learning. Data was used to train various object detection models ranging in accuracy from 65-99% IoU. Containerized Blender was run in AWS Batch service.
• Lead development of an Industrial Internet of Things project to collect data from 6 types of equipment across 4 lines producing breaded and fried product for food service customers. Project utilized AWS SiteWise to collect data at the edge, where it was sent to the cloud timeseries database and offloaded to S3 and made queriable with Athena.
• Lead program to develop the Data Egress Unified Compute Engine (DEUCE) for delivering curated data from a data warehouse built on Big Query. Solution provided an event driven service bus triggered on data changes and built with Google Cloud Platform Dataflow and PubSub for delivery of data downstream.
• Lead program to develop and deploy Google's Manufacturing Connect Edge solution built on Litmus OS for bi-directional data delivery for sending enterprise production order and product master data via DEUCE from cloud to edge and IIoT data from edge to cloud.
• Lead COVID Analytics rapid analytics response team and "war room" for urgent ad-hoc analytics needs for 1 year period during acute covid outbreak, which affected the meat processing industry in unique ways. Leadership needed data to help make decisions on where and how to shift production, e.g. from food service to retail and whether infection spread in surrounding areas warranted closing of production plants.
• Lead "Food Storm" rapid analytics response team and "war room" for urgent ad-hoc analytics needs for 4 month period post SAP S4 go-live to address rampant IDOC failures preventing recording of production, and the shipping and receiving of inventory across the entire supply and storage network.
• Lead analytics effort to build Automation Opportunity Analytics, an analytics data model bringing together production and equipment data from SAP, HR data from Workday, and health and safety and engineering standards data from in-house systems.
• Lead Business Simulation project instrumental in a flawless migration from legacy mainframe to SAP in 2015. Developed automated system to compare data between systems to ensure proper function of order to cash process.
• Lead development of Tyson Retail Analytics team, our company's largest data warehouse built on retail point of sale data at the day, store, item level for Walmart, Sam's, and Delhaize Food Lion. Solution included automation of data ingestion framework, landing, staging, and analytics data marts and models in MSSQL along with an order and event management application built in .NET framework, reporting in SSRS and a semantic layer SSAS Tabular model. Solution included predictive statisitcal learning models built with SSAS and SAS to predict sales.

• Tyson Foods People Leader Foundations
• Tyson Foods Next Generation Leadership
• Certified Scrum Master
• Stanford Certificate in Statistical Learning
• Coursera Certificates in Regression Models, Statistical Inference, Reproducible Research, Machine Learning, Exploratory Data Analysis, R Programming, Getting and Cleaning Data, Data Scientist’s Toolkit, Practical Machine Learning, Functional Programming Principles in Scala
• Tyson Foods Team Member of the Quarter
• Tyson Foods Leadership Bootcamp
• Tyson Foods Financial Concepts
• System Development Lifecycle

• University of Arkansas Academy of Computer Science and Computer Engineering (UAACSCE)
• Association for Computing Machinery (ACM)
• Project Management Institute (PMI)