Dylan Levsey#
Technical Consultant | Automation | AI & LLM Workflows
Technical Consultant focused on building scalable test automation, documentation systems, and LLM-powered workflows. Passionate about solving complex problems at the intersection of automation, AI, and software engineering. Background in physics research, data analysis, and statistical modeling.
Selected implementation notes on delivery workflows, architecture patterns, and testing methods.
Education, experience, skills, and publications.
Undergraduate research in astrophysics and cosmology.
Data analysis, visualization, technical writing, and software projects.
Background, contact information, and CV download.
QA & Playwright Test Automation#
I built a Playwright E2E suite around reusable helpers, shared configuration, and readable tests. Explore how it evolved, how its functions work, and how test steps and screenshots supported documentation and user-manual drafts.
Start here: Follow the move from recorded browser actions to reusable helpers, shared setup, and configuration-driven tests.
Playwright TypeScript E2E Testing
See how the helper architecture, agent-assisted test maintenance, and docs-as-code workflow fit together, with complete source examples and a project timeline.
Test Automation Reusable Helpers Docs as Code
Helper and test walkthroughs#
Functions explained — A guide to the reusable email, address, screenshot, and diagnostic functions, with links to their full implementations.
Email template helpers — Shared interactions across applications, optional template fields, and complete test examples using inline values or JSON configuration.
Address helpers — Primary and mailing address workflows, configurable variations, and examples of calling the helpers from a test.
Named screenshots — How screenshot names and output paths connect test captures to images in user-manual drafts.
Docs as code — How helper source became a developer reference through Sphinx, custom presentation, and deprecation guidance.
Proof of Address walkthrough — A complete 69-line test, with its imports, shared setup, helper calls, and screenshot outputs explained.
Debugging, agents, and documentation#
Selector diagnostics — How diagnostic logs exposed candidate controls and selector changes, with later use of Playwright MCP for investigation.
Agent workflow — The progression from browser recordings to user stories, helper-based tests, execution feedback, and repairs.
User-manual drafts — How stories, test steps, and named screenshots supplied material for manuals with human review.
Testing Results & Engineering Experiments#
Explore the test evidence, delivery pipeline, and focused experiments behind this work, from reproducible inputs and parallel execution to training a small model for documentation tasks.
Testing and Results — Reproducible exploratory inputs, baseline failure analysis, and reported testing and documentation output metrics.
Codegen-to-Documentation Pipeline — How recorded interactions, shared JSON configuration, test generation, execution checks, and publishing fit into a repeatable workflow.
LLM Fine-Tuning Experiment — Training a small model to generate C# XML documentation comments, with the training configuration, resource usage, and example outputs.
Parallel Scaling: Worker Count vs Pass Rate — A comparison of Playwright worker counts, pass rates, and execution times, with charts and run data.
Education & Mathematical Background#
B.S. in Physics · Northern Kentucky University · 2017–2021
Minors in Mathematics (applied mathematics focus) and Astronomy. I pursued substantially more mathematics and programming coursework than my degree required, including advanced undergraduate mathematics, statistical computing, and computational research.
Relevant Undergraduate Coursework — Calculus I–III, linear algebra, finite mathematics and matrices, differential equations and PDE methods, Introduction to Higher Mathematics, Group Theory, and Cryptology. Statistics coursework included Probability and Statistics I and Statistics with Simulation and Resampling using R.
Programming and Computational Research — Elementary Programming (INF 120), Object-Oriented Programming I (CSC 260), Mathematics Software Programming (CSC 270), and Python-based Astronomical Techniques (AST 310).
Coursework and Research Relevant to LLMs — Linear algebra, calculus, statistical simulation, and tensor and black-hole research as foundations for later work with models and scientific software.
Python Research with PHOEBE — Beginning in 2018, I learned an unfamiliar scientific Python package and used MCMC to model binary-star systems over approximately three years.
I earned a 3.0+ GPA across 60 credits of upper-level (300+) coursework while carrying 16–18 credits per fall and spring term from Fall 2018 through Spring 2021, contributing to multiple research projects, and working in university IT. View the full education summary and course record.
Professional Documentation Sites#
I designed, built, and maintained the documentation platforms, automation, publishing workflows, and site architecture shown below in collaboration with product and engineering stakeholders. Every site includes:
CI/CD deployment from source changes to published docs
First-pass PDF formatting via a custom scoring algorithm
Accessibility-focused structure for broad usability
Dual delivery as web docs and downloadable manuals
Custom docs architecture: The Enterprise and SOLO API sites are customized Sphinx implementations using
reStructuredText+MyST Markdown, custom directives/roles, and tailored theme/navigation behavior.
Case study — how I connected YAML/OpenAPI documentation, API validation, Playwright regression testing, generated test references, and Sphinx/LaTeX publishing through CI/CD.
The platform eventually supported structured documentation requests that fed an automated drafting and validation workflow. For one 70-page manual, the initial production cycle fell from approximately 1.5 months to under one hour, while retaining human review and quality gates.
Playwright OpenAPI Sphinx Python GitHub Actions LaTeX
Scope: Multi-role user manuals generated through a Codegen → QA → Documentation pipeline.
Impact: ~20,000% increase in documentation output.
CI/CD pipeline: Simultaneous HTML site and PDF manual build output from one pipeline run, keeping content synchronized across 4 manuals.
Delivery speed: Significantly reduced PDF-to-documentation-site turnaround time.
Customization: Custom Sphinx structure and authoring workflow using mixed
reStructuredTextandMyST Markdown, with project-specific styling and content patterns.Explore more: CIT docs home.
Focus: Endpoint-by-endpoint REST API reference.
Workflow: Automated generation with Python-based tooling.
Customization: Custom Sphinx + OpenAPI pipeline with mixed
reStructuredText/MyST Markdowncontent, tuned API navigation, and template-level customization.
Focus: Automated testing documentation for Playwright helpers and test utilities.
Customization: Local
sphinx-jsfork plus custom Sphinx directives for test reports, status tracking, and TypeScript helper docs.
Format: Step-by-step onboarding guide.
Access: Online documentation with downloadable PDF versions.
Writing Sample(s)#
Professional#
Teacher Created Materials | 2021–2022 | Target audience: 5th grade
180 Days of Reading, 2nd Edition (Teacher Created Materials, 2023),4 nonfiction units and fiction passages for 11 units.
View Writing Samples | View Publication
South America (Teacher Created Materials, 2023),Published book targeting a Flesch-Kincaid fifth-grade reading level.
Recommendations & Testimonials#
Former colleagues and managers have highlighted my ability to understand complex work, communicate across technical and operational teams, improve processes, and move shared work forward.
“His ability to comprehend complex subject matter and present it in a clear and concise manner has been instrumental in enhancing the overall user experience.”
Former colleague and Technical Writer · July 2023
From a written letter addressing research, documentation quality, cross-functional communication, troubleshooting, and process improvement.
“He has played a major role in getting us to where we are today, and his contributions have been meaningful and deeply appreciated.”
Former manager · Team-wide recognition
Shared during an organizational transition that reflected an onsite business need rather than my performance.
Additional highlights
“He has excelled in every way.”
“Dylan is welcome to any onsite position that becomes available at any time.”
“His work ethic, dedication, and integrity have earned him that standing.”
Full professional reference details are available upon request.