Clayton is available for hire
Hire ClaytonClayton Lemons
Verified Expert in Engineering
Software Developer
Ithaca, NY, United States
Toptal member since April 23, 2020
Bio
Clayton has spent 15 years building the platforms AI runs on and designing the security that gets them approved. He led the team behind Elevance Health’s AI/ML platform for data on 70 million patients, led cloud engineering on a FedRAMP High data and AI platform at the U.S. Department of Veterans Affairs through its authorization, and owned a 0-to-1 AI SaaS product on Azure. Kubernetes (CKA, CKAD, CKS), AWS, Azure, Terraform, Python, C#.
Portfolio
UnblockedOps
Security Architecture, Vulnerability Research, Enterprise Security Architecture...
Stealth AI-powered Legal Healthcare SaaS Startup (BlueFlag contract)
Python 3, PostgreSQL, Terraform, Azure, Azure CLI, Microsoft Entra ID, GitHub...
U.S. Department of Veterans Affairs
Python 3, PostgreSQL, Terraform, Azure, Azure CLI, Microsoft Entra ID, GitHub...
Experience
- Python - 15 years
- Security - 6 years
- Artificial Intelligence (AI) - 6 years
- Kubernetes - 4 years
- Amazon Web Services (AWS) - 4 years
- Azure - 3 years
- Machine Learning Operations (MLOps) - 3 years
- Large Language Models (LLMs) - 3 years
Work Experience
Founder
2026 - PRESENT
UnblockedOps
- Fractional security architect. I close the security findings that are holding something up, whether they came from an audit, a security review, or anywhere else. I design the fix, get it approved, and guide your engineers while they build it.
- Found and reported a Critical-severity information disclosure on a major US retailer's site through HackerOne. Confirmed it server-side; the vendor re-rated it from Medium to Critical, fixed it, and paid a $6K bounty (Sep 2026).
- Built the practice’s operating system on Claude Code: a chief-of-staff agent that plans and dispatches work on a GitHub issue board, specialist agents for research and writing, and a panel of reviewer agents that critique every draft before I see it.
Technologies: Security Architecture, Vulnerability Research, Enterprise Security Architecture, Certified Ethical Hacker (CEH), FedRAMP, HIPAA, Identity & Access Management (IAM), Claude, Claude Code, Anthropic, AI Agents, Agentic AI, Model Context Protocol (MCP), Python, TypeScript, Bash Script, GitHub, GitHub Actions, Astro, Cloudflare, PostHog, Security, AI Consulting, AI Security, AI Workflow, Claude API, Generative Artificial Intelligence (GenAI), Human-in-the-loop (HITL), Prompt Engineering, RAG Systems
Principal Full-stack Engineer
2024 - 2026
Stealth AI-powered Legal Healthcare SaaS Startup (BlueFlag contract)
- Owned architecture and delivery of the 0-to-1 product with a 3-person team across two time zones, and stayed hands-on with the code throughout.
- Designed the application as a .NET back end and an Angular front end on AKS, with Azure services behind them, and made the technical decisions from library and tool choices to database design and deployment strategy.
- Built the infrastructure as Terraform against Azure and AKS, so development, staging, and production came up from the same code and stayed identical.
- Packaged the application as a Helm chart with a values file per environment, so a deployment was a reviewable config change.
- Set up GitOps deployment with Argo CD so the team shipped on its own, several times a day, without waiting on me.
- Debugged and fixed critical production issues, including environment-specific runtime failures, keeping the platform stable and the team moving.
- Architected the AI agent (Semantic Kernel, Azure AI Foundry) that searches the platform for experts and matches them to an attorney’s case, so attorneys get a ranked shortlist instead of searching by hand.
Technologies: Python 3, PostgreSQL, Terraform, Azure, Azure CLI, Microsoft Entra ID, GitHub, GitHub Actions, Azure Kubernetes Service (AKS), Docker, Containerization, Azure Key Vault, Azure Blob Storage, Pytest, C#, .NET, Angular, Azure AI Foundry, Semantic Kernel (SK), Argo CD, GitOps, Helm, Single Sign-on (SSO), OAuth 2, Architecture, DevOps, API Design, Back-end, Claude, Large Language Models (LLMs), Artificial Intelligence (AI), Chatbots, Deployment, Prompt Engineering, Python, AI Agents, Recommendation Systems, Agentic AI, AI Architecture, AI Workflow, Embeddings from Language Models (ELMo), Generative Artificial Intelligence (GenAI), Hybrid Search, LLM Integration, Retrieval-augmented Generation (RAG), Semantic Search, Vector Search, Technical Leadership, Code Review
Cloud Engineering Lead
2024 - 2026
U.S. Department of Veterans Affairs
- Lead engineer for a 7-person team on the Summit Data Platform, a FedRAMP High data, analytics, and AI platform serving thousands of VA users.
- One of two engineers who took over the platform with little handover. Worked the security remediations on a 30/60/90-day plan, pulled down the service request backlog, and led the rollout of a half-deployed governance stack.
- Represented engineering in the platform’s audits, security reviews, questionnaires, and SOP tabletop exercises, and designed and implemented fixes and mitigations for the significant findings.
- Set up Argo CD and made GitOps the deployment standard on OpenShift, cutting release cycles from weeks to days and nearly eliminating configuration drift.
- Replaced manual, unversioned OpenShift deployments of the governance stack (Starburst, Immuta) with Helm charts and values files delivered through Argo CD, and managed secrets in Azure Key Vault via the External Secrets Operator.
- Wrote a Python SCIM provisioner that synced Entra ID users and groups into Starburst and Immuta when the identity provider could not reach the private network, unblocking the enterprise rollout and keeping identities consistent for Immuta’s ABAC.
- Led the technical onboarding of 20+ teams onto Starburst and Immuta, with weekly working sessions alongside the product owners and customer teams to clear blockers.
- Built the Azure Application Gateway path that let Databricks workspaces outside the VNet reach Starburst and Immuta: TIC 2.0 firewall routes for Microsoft-managed source ranges and a WAF allowlisting the Databricks NAT addresses.
- Managed the Azure and AWS infrastructure in Terraform: Application Gateway, Databricks, Synapse, SQL Server, PostgreSQL, virtual machines, and private endpoints.
- Automated deployment of containerized self-hosted integration runtimes on Docker Swarm, giving 200+ workgroups access to on-premises data through Azure Synapse and Data Factory.
Technologies: Python 3, PostgreSQL, Terraform, Azure, Azure CLI, Microsoft Entra ID, GitHub, GitHub Runners, GitHub Actions, Red Hat OpenShift, Amazon EKS, Azure Key Vault, Trino, Windows PowerShell, Active Directory (AD), Azure Databricks, Azure Synapse Analytics, Pytest, Kubernetes, Argo CD, GitOps, SCIM, Identity & Access Management (IAM), OAuth 2, Azure Application Gateway, ABAC, Docker Swarm, Containerization, Docker, Docker Compose, Dynatrace, Single Sign-on (SSO), Architecture, Artificial Intelligence (AI), Helm, Starburst, Immuta, Neo4j, Row-level Security (RLS), Role-based Access Control (RBAC), FedRAMP, Authority to Operate (ATO), Security Compliance, External Secrets Operator, Technical Leadership, Amazon Web Services (AWS), Microsoft SQL Server, Azure Data Factory (ADF), Web Application Firewall (WAF), Azure Virtual Networks, Security Architecture, Python, Incident Response, Data Pipelines, Database Migration, Best Practices, Pipelines, Security, AI Agents, Microsoft Copilot, Code Review
Partner
2024 - 2026
BlueFlag LLC
- Technical partner. Led delivery and owned the architecture decisions for two concurrent engagements: a 7-person team on a federal data and AI platform and a 3-person team on a 0-to-1 AI SaaS product.
- Built the engineering side of the firm’s transition from talent placement to cloud engineering and AI (infrastructure, agents, and applications), with the federal platform and the SaaS product as its first delivery engagements.
- Built the firm’s technical hiring process: designed and evaluated HackerRank assessments, ran the technical interviews, and onboarded the new engineers.
- Wrote the technical side of the firm’s proposals and statements of work.
- Set the firm’s engineering standards and ran its technical operations, including the GitHub organization, cloud tenants, access management, and AI coding tools for both teams.
- Mentored the engineers on both teams, from code review to growth goals.
- Stepped in more than once when upstream deliverables the federal client expected had stalled and the work landed on the firm, and got them finished; those recoveries helped keep the contract alive.
Technologies: Azure, Red Hat OpenShift, Kubernetes, Technical Leadership, Team Leadership, Remote Team Leadership, Architecture, Cross-functional Team Leadership, Solution Architecture, Cloud Architecture, Amazon Web Services (AWS), IT Strategy, Consulting, IT Consulting, Strategic Planning & Execution, Software Project Management, Business Requirements, Generative Artificial Intelligence (GenAI), Semantic Kernel (SK)
Senior Back-end Engineer
2024 - 2024
Farmstream LTD
- Designed and implemented a cost-optimized, serverless timelapse generation system for processing still images from thousands of farm cameras into streamable videos using AWS Batch, Step Functions, Lambda, and a containerized Python/FFMPEG script.
- Built a time-lapse system with end-to-end automation, including infrastructure provisioning using AWS CloudFormation and CI/CD using AWS Code pipeline and CodeBuild, securely integrated with GitHub via federated identity.
- Diagnosed and resolved a critical VPN connectivity issue blocking client access to customer cameras on the 4G provider network.
- Exceeded contract scope and client expectations by redesigning the VPC and subnet architecture as part of VPN connectivity resolution, establishing proper network segmentation (public, private, VPN-connected subnets) with secure resource isolation.
- Exceeded contract scope and client expectation by managing network resource provisioning using CloudFormation IaC.
- Enabled new product capabilities (camera streaming to web) that were previously blocked by VPN issues and network architecture limitations.
- Demonstrated full-stack cloud expertise and senior-level independent consultancy across a single engagement: requirements gathering, serverless architecture, video processing, network engineering, IaC, CI/CD, and security best practices..
Technologies: Amazon Web Services (AWS), APIs, AWS Lambda, Amazon S3 (AWS S3), SQL, Python 3, Python, Serverless, Serverless Architecture
Enterprise Strategy Architect
2023 - 2024
U.S. Department of Veterans Affairs
- Produced the technical strategy document for a large-scale, cloud-based data and analytics platform, helping stakeholders align on a governance framework, integration patterns, and analytics capabilities.
- Created a comprehensive test plan to validate a custom monitoring application for data integration pipelines in Azure Data Factory and Azure Synapse Analytics.
- Built a YAML-based declarative testing tool, enabling testers not familiar with Data Factory or Synapse to validate the application using simulated integration pipelines.
- Assumed a technical lead role when the lead engineer departed before a crucial deadline, quickly mastering C# to resolve project gaps and deliver on time, significantly contributing to an $80 million follow-on contract win.
- Identified scope reduction opportunities that saved several days of development effort on a tight deadline while preserving MVP functionality, enabling on-time delivery.
- Implemented back-end engine components for monitoring distributed data integration jobs.
- Improved codebase quality through systematic bug fixes, unit test coverage expansion by 85% for previously untested code, and implementation of software engineering best practices.
- Identified and remediated critical security vulnerabilities related to secret management, implementing Azure Key Vault integration with managed identity for secure secret access in all environments.
Technologies: Big Data, Databases, ETL Tools, Data Management, IT Strategy, Cloud, Distributed Systems, Azure, C#, .NET, Python 3, MSTest, React, Back-end, SQL Server 2012, GitHub Actions, Azure Key Vault, Architecture, Python, Technical Leadership, Code Review
Senior Director of AI Technology
2023 - 2023
Elevance Health
- Managed an 8-member engineering team skilled in AI/ML, data science, back-end API development, and DevOps, fostering a diverse technical environment.
- Mentored eight engineers in technical and leadership skills, enabling two engineers to achieve technical leadership roles.
- Oversaw the maintenance of a cloud-native, elastic AI/ML platform built on JupyterHub and a data pipeline that processed HL7 FHIR resources for over 70 million patients, handling billions of claims and clinical records.
- Headed a cross-functional team to develop a research-focused data science platform on Google Kubernetes Engine, ensuring secure access to deidentified data for over 70 million patients.
- Designed a secure solution using Kasm to prevent data exfiltration from container-based JupyterLab workspaces running on the data science platform.
- Containerized a customized JupyterLab environment for the data science platform equipped with an extensive set of AI/ML tools.
- Architected a secure, multi-tenant persistent storage system to support both user and shared project data in JupyterLab workspaces, employing Google Cloud Storage buckets mounted with gcsfuse.
- Engineered a deployment strategy for an innovative synthetic data generation system.
- Served as the cloud engineering lead for an LLM-powered, internal AI chatbot for CSR queries, providing solutions for both the LLM's development and its deployment and integration with the tool's front-end application.
- Implemented Ray on Kubernetes with autoscaling and GPU support and showcased its ability to fine-tune a 70 billion parameter Llama 2 LLM using the Fully Sharded Data Parallel (FSDP) technique.
Technologies: Code Review, Coaching, Career Coaching, Technical Leadership, Software Design, Software Architecture, Feedback Review, Agile, Scrum, Software Project Management, Kubernetes Operations (kOps), Director, Software Engineering, Cross-functional Team Leadership, Strategic Planning & Execution, Idea Synthesization and Application, Staff Management & Development, Goal Management, Project Coordination, Business Requirements, Open-source LLMs, Large Language Models (LLMs), New Product Development, Google Kubernetes Engine (GKE), Google Cloud Platform (GCP), Google Cloud Storage, JupyterLab, Jupyter Notebook, React, Python, FastAPI, Terraform, Cross-functional Collaboration, Datadog, Ray, PyTorch, Llama 2, Fine-tuning, Data-level Security, Data Exfiltration Prevention, Git, Cloud Deployment, Full-stack Development, Leadership, Team Leadership, Remote Team Leadership, Snowflake, Protegrity, Retrieval-augmented Generation (RAG), ChatGPT, Chatbots, Distributed Systems, Real-time Data, Web Development, High-performance Computing (HPC), Artificial Intelligence (AI), LangChain, AI Chatbots, Best Practices, Conversational AI, Knowledge Bases, AI Model Fine-tuning, Machine Learning Operations (MLOps), AI Architecture, Fully Sharded Data Parallelism (FSDP), Generative Artificial Intelligence (GenAI), Hugging Face Transformers, LLM Integration, RAG Systems, Synthetic Data Generation, Vector Search, Go, Architecture
AI Software Engineering Lead
2021 - 2023
Elevance Health
- Led the architecture and team efforts to build and deploy a cloud-native AI/ML platform on AWS for enhanced data science on scalable infrastructure, integrating technologies such as Kubernetes, JupyterHub, Spark, MLflow, Feast, and Pachyderm.
- Secured cloud accounts for the AI/ML platform deployment, working in close collaboration with internal committees and DevSecOps teams to ensure compliance and governance alignment.
- Spearheaded the deployment of JupyterHub on Amazon EKS in close collaboration with DevSecOps engineers, leveraging EFS, EBS, S3, and custom Docker images to meet specific user environment needs.
- Developed a single sign-on (SSO) solution integrating JupyterHub and MLflow via Auth0, facilitating a seamless authentication experience for users.
- Developed a high-performance computing solution for users of the AI platform, seamlessly integrating on-demand, scalable GPU resources and elastic, Kubernetes-hosted Spark jobs with JupyterHub.
- Proposed and contributed a security solution to Pachyderm, improving the security posture of its "JupyterLab Pachyderm Mount Extension" and making it easier for users to integrate.
- Influenced Pachyderm's development roadmap by identifying numerous performance issues and suggesting new features and optimizations, several of which were implemented.
- Conducted comprehensive training sessions for data scientists and engineers on the effective use of the AI/ML platform, enhancing team capabilities.
- Directed a cross-functional team on the development and operationalization of a predictive model for type 2 diabetes on the AI/ML platform, successfully advocating for the use of Feast and MLflow to achieve MLOps best practices.
- Led the architectural design and development of a new FHIR generation pipeline to replace an old one, cutting processing time from three weeks to 24 hours for 70 million patients and substantially lowering operational costs.
Technologies: JupyterLab, Jupyter Notebook, Spark, PySpark, MLflow, Feast, Pachyderm, Amazon EFS, Amazon EKS, Amazon Elastic Block Store (EBS), AWS ELB, AWS IAM, AWS Command Line Interface (CLI), Amazon S3 (AWS S3), AWS Auto Scaling, Python 3, Python, GPU Computing, Distributed Computing, Big Data, Machine Learning Operations (MLOps), Technical Leadership, Architecture, Software Design, Machine Learning, GitLab CI/CD, GitLab, Helm, Artificial Intelligence (AI), Single Sign-on (SSO), Artifactory, OCI Artifact Registry, Docker, Docker Hub, Docker Compose, DevOps, HL7 FHIR Standard, Optimization, Git, Cloud Deployment, API Integration, Team Leadership, Remote Team Leadership, Leadership, Protegrity, Snowflake, Distributed Systems, Real-time Data, High-performance Computing (HPC), AI Architecture, Data Pipelines, Best Practices, Security, Code Review
AI Solutions Engineer Executive Advisor
2020 - 2021
Anthem
- Conducted comprehensive research to identify Pachyderm as an enterprise-grade COTS software solution that met specific needs for pipeline orchestration, distributed processing, incremental processing, and data tracking and provenance.
- Drove the procurement process for Pachyderm, successfully navigating licensing, negotiations, and acquisition.
- Directed and assisted a team of cloud engineers with the deployment of Pachyderm within the enterprise's AWS cloud infrastructure, specifically leveraging Amazon EKS.
- Oversaw the operationalization of Pachyderm, establishing robust processes and best practices for building and executing large-scale data pipelines.
- Standardized and documented data engineering best practices for the organization.
- Designed an internal, patient-focused health trajectory data structure, significantly enhancing data scientists' ability to rapidly analyze data and develop AI-driven health models.
- Led the design and development of a Pachyderm pipeline that hydrated the above health trajectory data structure with data from over 70 million patients, encompassing 2+ TB of data.
Technologies: Spark, PySpark, Kubernetes, Docker, Amazon EKS, Python 3, Python, Apache Airflow, Pachyderm, Data Pipelines, CI/CD Pipelines, GitLab CI/CD, GitLab, Flake8, Pytest, pre-commit, Orchestration, Data Structures, Big Data, Data Science, Code Review, Debugging, Software Testing, Distributed Computing, Distributed Software, Negotiation, Procurement, COTS, Enterprise SaaS, Provenance, Data Lineage, Artificial Intelligence (AI), IT Security, Technical Leadership, Git, Cloud Deployment, Snowflake, Protegrity, Distributed Systems, Real-time Data, High-performance Computing (HPC), Performance Optimization, Pipelines, Best Practices
AI ETL Solutions Engineer via Toptal
2020 - 2020
Anthem AI - Telehealth/PIP
- Spearheaded the design and execution of a complex data pipeline that enables the seamless delivery of AI-driven health insights from an on-prem server to a cloud-based application.
- Engineered a Python-based API and storage framework for the storage and retrieval of AI-driven health insights on Amazon S3, utilizing compression, Base64 encoding, and indexing for flexibility and efficiency.
- Enhanced data science operations by providing expert ETL and ML pipeline engineering support to data scientists in the form of code reviews, debugging, pair programming, and performance optimization.
- Maintained on-prem ETL pipeline components built using Hive, PySpark, and Airflow.
- Advocated successfully to leadership for the transformation of an on-prem ETL pipeline to a cloud-native solution, leveraging Snowflake, Kubernetes, PySpark, and Pachyderm.
Technologies: Python, SQL, ETL, PySpark, Data Analytics, Data Modeling, Data Profiling, Spark, Apache Hive, Hadoop, Data Pipelines, GitLab CI/CD, CI/CD Pipelines, Apache Airflow, Kubernetes, Pachyderm, Snowflake, Amazon S3 (AWS S3), APIs, Frameworks, Storage, Artificial Intelligence (AI), Machine Learning, Big Data, PyTorch, Python 3, Code Review, Debugging, Pair Programming, Advisory, Amazon Web Services (AWS), AWS Step Functions, AWS Lambda, Bitbucket, Git, Cloud Deployment, Distributed Systems, High-performance Computing (HPC), Performance Optimization, Pipelines
Research Software Engineer (Machine Learning)
2018 - 2020
GrammaTech
- Accelerated the static and binary analysis of a large-scale codebase by implementing a data and ML pipeline using Python, MongoDB, and JavaScript.
- Reduced computational costs for a binary analysis program by implementing a pupil-style ML model with scikit-learn.
- Crafted data analysis techniques in Python to detect security issues in JavaScript functions, such as swapped callback and error arguments in higher-order continuation-style functions.
- Leveraged the Doc2Vec model to vectorize function call sites and definitions, streamlining the detection of swapped arguments through semantic similarity of parameter and argument names.
- Developed the back end of a feature for binary scanning in a SaaS binary analysis tool.
Technologies: JavaScript, GitLab, Docker, C++, MongoDB, Keras, TensorFlow, Python, Scikit-learn, Machine Learning, Statistical Analysis, Bash Script, GitLab CI/CD, CI/CD Pipelines, NumPy, Big Data, Parallel Programming, Git, High-performance Computing (HPC), Security, Embeddings from Language Models (ELMo), Semantic Search, Code Review
Software Engineer I – III
2013 - 2020
National Instruments
- Earned recognition for outstanding performance, receiving the "Rookie of the Year" award, multiple fast-track promotions, and the opportunity to lead a key project.
- Standardized and streamlined the firmware downloading framework across multiple devices in the NI-DCPower and NI-DMM product families.
- Designed and implemented over 20 features in the NI-DCPower and NI-DMM driver APIs for Windows, with many improvements directly visible to the end user.
- Led the research and definition of three major features for a key product, collaborating with project managers across hardware teams and various stakeholders to ensure alignment and address technical requirements comprehensively.
- Designed and led the development of an internal programming language and compiler that targeted a proprietary instruction set for power supply output control, enabling flexible device behavior reconfiguration and complex output control.
- Implemented a client-server system that enables remote management of NI-DCPower API driver sessions, facilitating debugging and introspection.
- Improved developer workflow efficiency by implementing a Sublime Text plugin to integrate Perforce.
- Developed a VS Code extension that integrates NI's custom build system with Microsoft's C/C++ extension, enabling advanced features like semantic code completion.
- Researched the Tarantula fault localization technique, successfully created a prototype for select NI codebases, and showcased the findings at an internal engineering conference.
- Mentored more than 10 interns and junior engineers.
Technologies: Compiler Design, API Design, Firmware, Development, Windows, Visual Studio, Python, C++, Perforce, Windows Kernel Drivers, TypeScript, Sublime Text 3, Visual Studio Code (VS Code), Leadership, Team Leadership, Best Practices, Performance Optimization, Code Generators, Technical Leadership, Code Review
Web Developer
2012 - 2013
CleanTelligent Software
- Optimized several database queries and storage layouts, including the file storage system for customer photos, which reduced several API response times to just milliseconds.
- Implemented the back-end API for a customizable report generation tool.
- Applied a new UI theme to several parts of the website.
Technologies: Apache Struts 2, HTML, CSS, JavaScript, Jakarta Server Pages (JSP), Java, PostgreSQL, CVS, Full-stack Development, Web Development
Software Engineer Intern
2012 - 2012
National Instruments
- Designed an essential kernel driver feature that streamlined driver communication with embedded storage devices on over 10 commercial products.
- Developed a code generation tool to support the driver feature, which automatically leveraged Python and Mako templates to generate C++ and LabVIEW code from metadata.
- Investigated and presented the advantages and disadvantages of various metadata schema formats for the code generation tool, then led a consensus meeting to select the most suitable one.
Technologies: LabVIEW, Ruby, C++, Python, Mako, Ruby ERB, Perforce, Windows Kernel Drivers, Code Generators
Web Developer
2011 - 2012
CleanTelligent Software
- Pinpointed and documented multiple user experience inconsistencies across related functionalities and addressed the corresponding issues successfully.
- Broadened the capabilities of a crucial job scheduling tool by integrating additional back-end queries and introducing new user interface elements on the front end.
- Resolved over 30 bugs throughout the CleanTelligent website's front end and back end, enhancing overall performance and user experience.
Technologies: JavaScript, Apache Struts 2, HTML, CSS, Jakarta Server Pages (JSP), Java, PostgreSQL, CVS, Full-stack Development, Web Development
Project History
Ithaca Community Diaper Drive
http://ithacadiaperdrive.siteI built the website for the Ithaca Community Diaper Drive, pro bono. Diapers are in high demand in the Ithaca area, and the site has helped the drive gather over 6,000 diapers in two months, nearly 75% of its goal.
The site handles every part of the drive: showing progress and which sizes are filled or still needed, linking to the drive's Walmart and Amazon registries, listing drop-off locations, and receiving donors' notes of encouragement to families. Behind a login, volunteer admins configure the drive and record donations.
AI Support: Admins can request Claude-generated, versioned registry scripts for each retailer. The scripts parse prices nightly so donors can get the most for their money. When a volunteer scans an unknown barcode, Claude identifies the product's name, size, and diaper count. When donor notes are submitted, Claude filters out inappropriate ones before volunteers review the rest.
Security: Public forms are rate-limited. Claude is guarded against prompt injection. Volunteers are assigned RBAC roles. Outbound API requests are allowlisted. Admin changes are logged.
CI/CD: Over 900 unit and 250 e2e tests run through CI. Vercel automates deployments and database migrations.
The site handles every part of the drive: showing progress and which sizes are filled or still needed, linking to the drive's Walmart and Amazon registries, listing drop-off locations, and receiving donors' notes of encouragement to families. Behind a login, volunteer admins configure the drive and record donations.
AI Support: Admins can request Claude-generated, versioned registry scripts for each retailer. The scripts parse prices nightly so donors can get the most for their money. When a volunteer scans an unknown barcode, Claude identifies the product's name, size, and diaper count. When donor notes are submitted, Claude filters out inappropriate ones before volunteers review the rest.
Security: Public forms are rate-limited. Claude is guarded against prompt injection. Volunteers are assigned RBAC roles. Outbound API requests are allowlisted. Admin changes are logged.
CI/CD: Over 900 unit and 250 e2e tests run through CI. Vercel automates deployments and database migrations.
Copilot MCP Bridge for Self-Hosted Jira and Confluence
Engineers and managers on a federal platform team wanted AI help with their Jira tickets and Confluence pages, but the self-hosted Atlassian services had no AI features or MCP servers, and users couldn't generate API tokens. Since GitHub Copilot was already approved for use, I built a VS Code extension that gave Copilot access to both services, plus local Office files.
The extension runs three local MCP servers over stdio, so no network port is ever opened: two for the Atlassian services and one that reads Word, Excel, and PowerPoint files using officeParser. Agent skill files teach Copilot when to use each server and how to combine them.
Each Atlassian server connects over the respective service's REST API. To authenticate, Copilot runs an MCP tool that opens Chrome with the Chrome DevTools Protocol (CDP) enabled, waits while the user signs in through SSO as usual, and picks up the session cookies through CDP. Every subsequent tool call then authenticates through the cookies, allowing Copilot to act with exactly that user's permissions until the session expires.
The extension enabled users to answer questions across tickets, pages and documents, analyze past work, find and manage work items, and turn tickets into code.
The extension runs three local MCP servers over stdio, so no network port is ever opened: two for the Atlassian services and one that reads Word, Excel, and PowerPoint files using officeParser. Agent skill files teach Copilot when to use each server and how to combine them.
Each Atlassian server connects over the respective service's REST API. To authenticate, Copilot runs an MCP tool that opens Chrome with the Chrome DevTools Protocol (CDP) enabled, waits while the user signs in through SSO as usual, and picks up the session cookies through CDP. Every subsequent tool call then authenticates through the cookies, allowing Copilot to act with exactly that user's permissions until the session expires.
The extension enabled users to answer questions across tickets, pages and documents, analyze past work, find and manage work items, and turn tickets into code.
Lemons House
https://lemons.houseI built Lemons House, a modular household management app my family of eight runs our home on.
The first module (chores) includes: a chore library with illustrated instructions; schedules that create each day's tasks; rotations that decide whose turn it is; effort and time set by age; kid-suggested chore edits parents approve; parental review flows; daily reminders and notifications.
AI Support: A Claude assistant on a custom streaming agent loop answers chore and task questions, builds reports from plain English, and drafts new chores. It reads through the app's API with the member's permissions, and runs a change only after they confirm it. A golden-set eval gates prompt and tool changes.
Security: Every query is scoped to the member's role. Assistant controls map to the OWASP LLM Top 10: inputs are escaped, outputs render as plain text, and use is capped per member and family. Rich text is sanitized, redirects are allowlisted, and the server fetches no user- or model-supplied URL.
Production: Merges to main deploy to Vercel and migrate the database. Each branch gets a preview with its own database. Over 1,900 unit and 80 e2e tests run through CI, and every pull request requires a Claude code and security review.
The first module (chores) includes: a chore library with illustrated instructions; schedules that create each day's tasks; rotations that decide whose turn it is; effort and time set by age; kid-suggested chore edits parents approve; parental review flows; daily reminders and notifications.
AI Support: A Claude assistant on a custom streaming agent loop answers chore and task questions, builds reports from plain English, and drafts new chores. It reads through the app's API with the member's permissions, and runs a change only after they confirm it. A golden-set eval gates prompt and tool changes.
Security: Every query is scoped to the member's role. Assistant controls map to the OWASP LLM Top 10: inputs are escaped, outputs render as plain text, and use is capped per member and family. Rich text is sanitized, redirects are allowlisted, and the server fetches no user- or model-supplied URL.
Production: Merges to main deploy to Vercel and migrate the database. Each branch gets a preview with its own database. Over 1,900 unit and 80 e2e tests run through CI, and every pull request requires a Claude code and security review.
Genstrat
https://github.com/claytonlemons/genstratA random train/test split can leave rare entity types thin or missing from the test set, which makes an NER model's scores misleading. Genstrat finds a better split with a genetic algorithm that scores each candidate by how closely the train and test sets match, class by class, using Jensen-Shannon divergence. Every class counts equally, so rare entity types carry the same weight as common ones.
Scoring every candidate in every generation was the bottleneck, so I wrote the fitness function as a CUDA kernel and packaged it as a Python extension that drops into PyGAD, with a CPU fallback for machines without a GPU.
The kernel runs one block per candidate split, keeps the class counts resident in GPU memory across evaluations, builds histograms in shared memory, and sums the divergence terms with a per-class tree reduction. Only the candidate splits and their scores cross the PCIe bus.
Against an optimized NumPy/SciPy baseline, it runs about 100x faster on realistic workloads (500 candidates, 100,000 samples) and 20 to 40x faster on smaller batches. C++ and Python test suites and a GPU benchmark ship with it.
Scoring every candidate in every generation was the bottleneck, so I wrote the fitness function as a CUDA kernel and packaged it as a Python extension that drops into PyGAD, with a CPU fallback for machines without a GPU.
The kernel runs one block per candidate split, keeps the class counts resident in GPU memory across evaluations, builds histograms in shared memory, and sums the divergence terms with a per-class tree reduction. Only the candidate splits and their scores cross the PCIe bus.
Against an optimized NumPy/SciPy baseline, it runs about 100x faster on realistic workloads (500 candidates, 100,000 samples) and 20 to 40x faster on smaller batches. C++ and Python test suites and a GPU benchmark ship with it.
NI-DCPower Soft Front Panel Debug
https://ni.com/pdf/manuals/375374a.htmlNI-DCPower 14.1 shipped a way to debug a power supply or source measure unit while another program is using it: an engineer opens the Soft Front Panel and monitors or controls the live session. I built the system underneath that feature.
A DLL inside the driver session starts an Apache Thrift server on localhost, and a Python/C++ client mirrors the driver's API, acting as a proxy into the running session from a separate process. I chose Thrift so the same channel could later accept connections from other machines.
LabVIEW and C# developers built the Soft Front Panel feature on my client API, and I worked with them through development so the API fit how they used it.
A DLL inside the driver session starts an Apache Thrift server on localhost, and a Python/C++ client mirrors the driver's API, acting as a proxy into the running session from a separate process. I chose Thrift so the same channel could later accept connections from other machines.
LabVIEW and C# developers built the Soft Front Panel feature on my client API, and I worked with them through development so the API fit how they used it.
Subforce
https://github.com/claytonlemons/SubforceMy teammates at National Instruments preferred Sublime Text, but our code lived in Perforce, so every checkout, diff, or submit meant switching tools. I built Subforce, an open-source Sublime Text plugin that brings 19 Perforce commands into the editor, including sync, checkout, revert, changelist management, graphical diffs, and automatic checkout on save.
Because Sublime Text runs plugins in its own embedded Python, I also packaged Perforce's P4Python API as a separate Sublime dependency, and worked with the Package Control community to publish the plugin there.
Because Sublime Text runs plugins in its own embedded Python, I also packaged Perforce's P4Python API as a separate Sublime dependency, and worked with the Package Control community to publish the plugin there.
Python SCIM Provisioner for Private Azure Data Platform
A federal data platform on Azure ran its governance tools, Starburst and Immuta, inside a private virtual network. Rollout, onboarding, and RBAC/ABAC policy configuration were all blocked because users and groups hadn't been provisioned due to technical and security limitations. Entra ID's built-in SCIM provisioner couldn't reach the tools without opening a public endpoint into the network, and it didn't support nested groups or database-sourced custom attributes needed by the Immuta ABAC policies.
I built a SCIM provisioning job in Python that runs inside the network as a containerized Kubernetes CronJob, alongside the Starburst and Immuta deployments. It pulls users and group memberships from Microsoft Graph, flattens nested groups, adds the database-sourced attributes, and pushes the result to each tool through its SCIM API, with the calls running concurrently under asyncio. Entra ID stayed the single source of truth for identity, as the platform's security architecture required.
The job unblocked the governance rollout across the platform, and I went on to lead the technical onboarding of more than 20 teams onto Starburst and Immuta.
I built a SCIM provisioning job in Python that runs inside the network as a containerized Kubernetes CronJob, alongside the Starburst and Immuta deployments. It pulls users and group memberships from Microsoft Graph, flattens nested groups, adds the database-sourced attributes, and pushes the result to each tool through its SCIM API, with the calls running concurrently under asyncio. Entra ID stayed the single source of truth for identity, as the platform's security architecture required.
The job unblocked the governance rollout across the platform, and I went on to lead the technical onboarding of more than 20 teams onto Starburst and Immuta.
Education
2014 - 2016
Master of Science Degree in Software Engineering
The University of Texas at Austin - Austin, TX, USA
2007 - 2013
Minor in Mathematics
Brigham Young University - Provo, UT
2007 - 2013
Bachelor of Science Degree in Computer Science
Brigham Young University - Provo, UT, USA
Certifications
JANUARY 2026 - PRESENT
Introduction to Parallel Programming with CUDA
Johns Hopkins University | via Coursera
JANUARY 2026 - PRESENT
Introduction to Concurrent Programming with GPUs
Johns Hopkins University | via Coursera
JANUARY 2026 - JANUARY 2028
[NCA-AIIO] NVIDIA-Certified Associate in AI Infrastructure and Operations
NVIDIA
DECEMBER 2025 - PRESENT
Getting Started with Accelerated Computing in Modern CUDA C++
NVIDIA
DECEMBER 2025 - PRESENT
Fundamentals of Accelerated Computing with CUDA Python
NVIDIA
AUGUST 2024 - PRESENT
Managed Services on AWS and DevOps
Great Learning
AUGUST 2024 - PRESENT
Cloud Computing on AWS
Great Learning
AUGUST 2024 - PRESENT
Cloud Foundations
Great Learning
AUGUST 2024 - PRESENT
Managed Services on Azure
Great Learning
AUGUST 2024 - PRESENT
Azure Essentials
Great Learning
FEBRUARY 2024 - FEBRUARY 2026
CKS: Certified Kubernetes Security Specialist
The Linux Foundation
NOVEMBER 2023 - NOVEMBER 2026
CKA: Certified Kubernetes Administrator
The Linux Foundation
OCTOBER 2023 - OCTOBER 2026
CKAD: Certified Kubernetes Application Developer
The Linux Foundation
DECEMBER 2019 - PRESENT
Parallel Programming in Scala
Coursera
DECEMBER 2019 - PRESENT
Functional Programming in Scala Capstone
Coursera
DECEMBER 2019 - PRESENT
Functional Programming Principles in Scala
Coursera
DECEMBER 2019 - PRESENT
Functional Program Design in Scala
Coursera
DECEMBER 2019 - PRESENT
Big Data Analysis with Scala and Spark
Coursera
APRIL 2019 - PRESENT
Advanced Python
Skills
Libraries/APIs
TensorFlow, Keras, NumPy, SciPy, Pandas, Scikit-learn, PySpark, Claude API, Asyncio, React, PyTorch, Ruby ERB, Backbone.js, Mustache, Windows API, Azure Cognitive Services, Azure API Management, Node.js, Playwright, REST APIs, Jira REST API, Hugging Face Transformers
Tools
Sublime Text 3, Visual Studio, Gensim, Git, Perforce, GitLab, GitLab CI/CD, Claude, Claude Code, LabVIEW, Apache Airflow, Pachyderm, Amazon EKS, Pytest, Helm, Logging, AWS Step Functions, Bitbucket, Amazon Elastic Block Store (EBS), AWS ELB, AWS IAM, AWS Command Line Interface (CLI), Artifactory, Docker Hub, Docker Compose, Google Kubernetes Engine (GKE), Terraform, CVS, ChatGPT, GitHub, Azure Kubernetes Service (AKS), Azure Key Vault, Azure Application Gateway, Docker Swarm, Microsoft Copilot, Dynatrace, MSTest, Azure Logic Apps, NVIDIA Nsight Systems, Kubernetes Operators, NVIDIA Jetson, Named-entity Recognition (NER), Azure Machine Learning, Amazon Virtual Private Cloud (VPC), Amazon Elastic Container Service (ECS), AWS CloudTrail, AWS CodeCommit, AWS CodeBuild, AWS CodeDeploy, AWS CloudFormation, Amazon ElastiCache, AWS Cost Explorer, Azure Monitor, Azure Network Security Groups, Astro, Prisma, Vitest, GitHub Copilot, Chrome Developer Tools, Jira, Confluence, Drizzle ORM
Languages
Python, C++, Bash, TypeScript, PHP, CSS, Ruby, Java, Scala, GraphQL, HTML, JavaScript, Bytecode, Python 3, Embedded C, C, Bash Script, SQL, Snowflake, C#, C++17, Go
Paradigms
Best Practices, Functional Programming, Concurrent Programming, Distributed Computing, ABAC, High-performance Computing (HPC), Serverless Architecture, Model Context Protocol (MCP), Role-based Access Control (RBAC), Compiler Design, Software Testing, Parallel Programming, ETL, Pair Programming, DevOps, HL7 FHIR Standard, Agile, Scrum, Parallel Computing, Azure DevOps, Synthetic Data Generation
Platforms
Kubernetes, Docker, Amazon Web Services (AWS), Cloud Native, Azure, NVIDIA CUDA, Windows, NetBeans, Software Design Patterns, Linux, AWS Lambda, Jupyter Notebook, OCI Artifact Registry, Director, Google Cloud Platform (GCP), Visual Studio Code (VS Code), Red Hat OpenShift, Azure Synapse Analytics, NVIDIA NeMo, Azure Event Hubs, Azure IaaS, Azure PaaS, Azure Data Lake Storage, Amazon EC2, AWS Elastic Beanstalk, Azure Functions, PostHog, Vercel
Frameworks
Spark, Flask, .NET, Hadoop, Jakarta Server Pages (JSP), Apache Struts 2, CakePHP, Ray, Bootstrap, Apache Thrift, Trino, Windows PowerShell, OAuth 2, Angular, Next.js, Django
Storage
PostgreSQL, MongoDB, Data Pipelines, Cloud Deployment, Databases, Apache Hive, Amazon S3 (AWS S3), Amazon EFS, Google Cloud Storage, Datadog, Microsoft Entra ID, SQL Server 2012, Azure SQL, Azure Cosmos DB, AWS Storage Gateway, Neo4j, Microsoft SQL Server, Database Migration
Other
Windows Kernel Drivers, Software Architecture, Back-end, Artificial Intelligence (AI), Code Review, Technical Leadership, Architecture, Leadership, Distributed Systems, Web Development, AI Architecture, Code Generators, Generative Artificial Intelligence (GenAI), Machine Learning, Security, Machine Learning Operations (MLOps), Cloud, GPU Computing, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Real-time Data, Active Directory (AD), Prompt Engineering, AI Agents, Agentic AI, AI Security, LLM Evaluation, Performance Optimization, Serverless, Conversational AI, Benchmarking, Pipelines, AI Workflow, Anthropic SDK, Embeddings from Language Models (ELMo), Function Calling, Human-in-the-loop (HITL), LLM Integration, RAG Systems, Development, Full-stack, Firmware, API Design, Concurrent Computing, Thread Scheduling, Processing & Threading, Code Validation, Data Mining, Software Project Management, Data Engineering, Programming, Operating Systems, Data Structures, Software Design, Asymmetric Encryption, Server Development, Non-blocking I/O, Natural Language Processing (NLP), Compilers, Digital Signal Processing, Computer Engineering, Algorithms, Statistical Analysis, CI/CD Pipelines, Big Data, Data Analytics, Data Modeling, Data Profiling, APIs, Frameworks, Storage, Debugging, Advisory, Flake8, pre-commit, Orchestration, Data Science, Distributed Software, Negotiation, Procurement, COTS, Enterprise SaaS, Provenance, Data Lineage, IT Security, Functional Design, Kubernetes Operations (kOps), Open Source, Container Orchestration, Troubleshooting, Scheduling, Site Reliability Engineering (SRE), System Administration, Containerization, JupyterLab, MLflow, Feast, AWS Auto Scaling, Single Sign-on (SSO), Optimization, Coaching, Career Coaching, Feedback Review, Software Engineering, Cross-functional Team Leadership, Strategic Planning & Execution, Idea Synthesization and Application, Staff Management & Development, Goal Management, Project Coordination, Business Requirements, Open-source LLMs, New Product Development, FastAPI, Cross-functional Collaboration, Llama 2, Fine-tuning, Data-level Security, Data Exfiltration Prevention, Mako, Version Control, Plugin Development, Client-server Model, DLL, Device Drivers, Full-stack Development, API Integration, ETL Tools, Data Management, IT Strategy, Team Leadership, Remote Team Leadership, Protegrity, Chatbots, Azure CLI, GitHub Runners, GitHub Actions, Azure Databricks, Argo CD, GitOps, Semantic Kernel (SK), SCIM, Identity & Access Management (IAM), Azure AI Foundry, Azure Blob Storage, CUDA Kernel, Numba, Graphics Processing Unit (GPU), NVIDIA vGPU, NVIDIA Triton, NVIDIA TensorRT, Deep Learning, NVIDIA GPU Operator, NVIDIA NGC, NVIDIA Isaac, NVIDIA NVLink, NVIDIA AI Enterprise, NVIDIA MIG, NVIDIA NIM, NVIDIA DGX, Slurm Workload Manager, pygad, Genetic Algorithms, Python Extensions, Linear Algebra, Calculus, Number Theory, Cloud Computing, Microsoft Entra, Azure Service Bus, Azure Event Grid, Azure Virtual Machines, Load Balancers, AWS DevOps, AWS CodePipeline, AWS NAT Gateway, Amazon Route 53, Azure AD Connect, IT Governance, Azure Resource Manager (ARM), Azure Virtual Networks, Deployment, LangChain, AI Chatbots, Security Architecture, Vulnerability Research, Enterprise Security Architecture, Certified Ethical Hacker (CEH), FedRAMP, HIPAA, Anthropic, Cloudflare, Solution Architecture, Cloud Architecture, Consulting, IT Consulting, Starburst, Immuta, Row-level Security (RLS), Authority to Operate (ATO), Security Compliance, External Secrets Operator, Azure Data Factory (ADF), Web Application Firewall (WAF), Front-end, Web Scraping, Extensions, Chrome DevTools, Chrome DevTools Protocol, Confluence API, Authentication, Session Handling, Application Security, OWASP Top 10, Server Sent Events (SSE), OWASP, RPC, IPC, Microsoft Graph API, Network Security, Data Governance, Incident Response, Knowledge Bases, Recommendation Systems, AI Model Fine-tuning, AI Consulting, Content Moderation, Fully Sharded Data Parallelism (FSDP), Hybrid Search, Semantic Search, Vector Search
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring