Devin Bost Email & Phone Number

Orem, Ut, Us

Spring, Tx

Santa Clara, California, Us

Building successful production generative AI pipelines, recommendation systems, and deep learning architectures for Fortune 500 customers to help them meet customer needs and generate new revenue streams.- Developed embedding model that demonstrated 50% improvement in retrieval performance (NDCG, AUROC, MAP, MRR) with contrastive learning, using PyTorch Lightning- Developed node2vec vector search notebook- Developed collaborative filtering vector search notebook- Developed semi-autonomous personalization framework for building RAG applications- Authored enterprise RFIs- Developed knowledge graph completion embedding model

Santa Clara, California, Us

Enabling companies to solve their greatest pain points at the root.- On Generative AI PoC SWAT Team. - Contributor to LangStream.- Contributor to Apache Pulsar.- Advocate of streaming best practices.- Developed numerous streaming applications and POCs for customers, such as:-- field level encryption-- CDC-- automatic Avro type generation-- streaming filters, transforms, in-memory deduplication-- migrations from on-prem to cloud-- OKTA / OAuth2 integrations, as well as with Keycloak

Orem, Ut, Us

Tech-savvy real estate holding company with properties in multiple states. We buy and rehab distressed properties.- For example: ran numbers, purchased distressed property at discount, organized team, rehabbed, and turned over a space-themed vacation rental in Oklahoma 100% remotely, turning a record profit and achieving "Guest favorite" on Airbnb without ever setting foot in the property

Midvale, Ut, Us

- Authored internal RFC standard for distributed tracing with OpenTelemetry- Authored internal RFC on JSON Schema for schema validation - Contributed to internal RFC on data fabric messaging architecture - Represented Overstock in community working groups, such as for Apache Pulsar, JSON Schema, and CloudEvents - Developed best practices, messaging patterns, and vision for Kappa architecture - Developed roadmap for leveraging Kubernetes and Gitops with ArgoCD and Tekton to modernize infrastructure - Educated stakeholders with *engaging* presentations- Created roadmap for adopting data mesh strategy to advance domain driven development across the company Accomplishments: - Conducted unsolicited company-wide survey on pain points and interviewed over 40 people across departments (including department execs) to identify root cause of pain points. Used that information to align stakeholders on developing a standards-based solution to unblock a multi-billion dollar opportunity involving real-time data and machine learning. - Developed enterprise streaming data fabric on Apache Pulsar to enable machine learning, search, reporting, and analytics to increase organizational ROI, enabling fully automated onboarding, provisioning, and visualization of hundreds of projects with thousands of topics and billions of messages per day - Worked closely with partners to resolve critical production bugs and optimize systems, reducing on-call incidents by 99.8%- Engineered tracing system for a messaging architecture with Jaeger and Pulsar using the OpenTracing standard- Contributed Go Function feature to Apache Pulsar using gRPC for distributed service health-check- Developed numerous applications and flows in Java, Python, and Golang to support data fusion and automation - Maintained and created migration plan to Kubernetes, Helm, and Terraform for clusters in Centos / Docker that were using SaltStack, Python, and bash in hybrid GCP/on-premise architectures

- Led research and development of machine learning and artificial intelligence and architectural implementations for production use. - Drove development of big data technology to accelerate the process of data-driven decision making. - Collected data and developed automated batch and streaming services to provide fast access to answers to critical questions to better meet the needs of customers.- Developed PySpark migration framework to ETL any database while automatically handling schema changes- Completed evaluations for Hive, Neptune, HBase, SQL Server, Mongo, Dynamo, Oracle

- Researched and developed artificial intelligence to teach students how to solve math problems without giving them direct answers- Led ML team to build abuse filter to block inappropriate student conduct- Led data science research and development team to design and test experiments and discover high-impact areas of improvement.- Developed proactive intervention architecture to detect and initiate chat with students at risk in real time, built CI/CD pipeline for it, and got it into production in 2 weeks

- Redesigned decision framework that decided what activities to give students next in our educational technology platform - Researched Markov decision processes, finite state automata, Bayesian models and neural networks, adaptive learning, and artificial intelligence applied to game design. - Implemented auto-mocking, dependency injection, and numerous object oriented design patterns to improve maintainability of sequencing framework - Researched graph methods in game learning - Researched use of supervised and unsupervised machine learning methods for personalizing content sequencing recommendations to students' individual learning needs to provide more adaptive content-recommendations and accelerate students' learning rates, including sequence to sequence neural networks and speech recognition models - Developed application versioning architecture

- Developed novel reduction method for reducing complexity of maximal clique (NP Hard) optimization problem via Katz Centrality- Built end-to-end enterprise web-applications architecture and applied artificial intelligence.- Researched maximal clique problem- Graph theory- Constraint Satisfaction Problems- Traveling Salesman Problem- Vertex Cover, 4 color problem, many other graph problems- Cook-levin theorem

Developed high-performance, elastic, high-security, cloud-based enterprise web applications using AngularJS, Node.js, UmbracoCMS, and Microsoft SQL Server.Developed solutions in WPF using Model-View View-Model (MVVM) with C#, and developed solutions with heavy use of T-SQL in high-volume transaction environments. Developed and deployed database driven web applications using HTML5, CSS3, ASP.NET 4.5 (MVC 4), JavaScript, JQuery, XPath, and XSLT.Wrote automation tasks using SQL Server Agent and PowerShell 3.0.Leveraged Test-Driven Development (TDD) on .NET with NUnit and xUnit. Designed Team Foundation Server (TFS) workflows in TFS 2013 with SQL Reporting Services (SSRS), SQL Server Analysis Services (SSAS), and SharePoint 2013. Configured managed networks with the IEEE 802.1Q standard. Implemented software design patterns and industry best practices.Developed RESTful web applications using Entity Framework 6.0 and WebAPI 2.Designed website user interfaces (UIs).Tested server-side and client-side JavaScript and JQuery with node.js.

I designed the network infrastructure to provide web services (across the WAN) in a high security environment with Team Foundation Server 2013, SharePoint 2013, and SQL Server Reporting Services (Enterprise 2012). I configured the network security hardware with 802.1Q VLAN trunking to isolate traffic flow and provide secure routing between virtual machines in the server farm. In both C# and VB.NET, I refactored user interfaces with these frameworks:1. Infragistics NetAdvantage 2013.12. Atalasoft DotImage 10.0My development work also included:1. Designing CLR database functions in C#, 2. reflection and meta-data driven T-SQL migration scripts (including dynamic stored procedures), 3. modeling use cases and code in Unified Modeling Language (UML),4. training software engineers to use .NET 4.5 techniques,5. integrating object-oriented design patterns into legacy software architecture,6. refactoring stored procedure design on Microsoft SQL Server 2008 R2 and 2012 Enterprise,7. training network engineers for adopting and implementing IEEE 802.1Q (VLAN trunking),8. deprecating VB6 code,9. planning migration to WPF (in C#) with the Model-View View-Model (MVVM) design pattern,10. utilizing Red Gate SQL Toolbelt tools,11. developing decentralized workflow with agile revision control (e.g. SCRUM) using Team Foundation Server (TFS) 2012, SharePoint (SP) 2013, and SQL Server Reporting Services (SSRS) 2012.12. planning, deploying, and configuring TFS 2012, SSRS 2012, and SP 2013 web services in a public environment, using best security practices.Additional aspect-oriented development techniques involved:1. Design of dynamic stored procedure invocation engine (from concept to final implementation) using Entity Framework 6.0 and ILGenerator using Expression Trees and principles of lambda calculus.2. Development of dynamic predicate builder for constructing data filters at runtime.

Computer Science

Education record