Robert MacDonald

Automation and AI Systems Architect: production platforms for pay and HR

Profile

I design and build automation and AI systems that are dependable enough to run every day: they check their own results, pick up where they left off after a failure, and keep a record of everything they did. These are not one-off tools. On my own time it has meant more than 90 independent software projects, from AI services to a 3D game engine.

I use AI coding assistants to build quickly, and I design the structure around them so that speed never comes at the cost of quality. Every change is checked automatically, every past mistake has a test that stops it from coming back, and the systems say plainly when they do not know something instead of guessing. And the systems are built to work together rather than stand alone: shared speech and AI services serve several of my applications, one router spreads work across my machines, and an index of everything I have built is open to AI assistants in any project, so each new piece of work starts from what already exists.

I work at the frontier: I evaluate new technology, prove what it can do in a real setting, then set it up and put it into everyday use. AI document processing, secure cloud environments and automation, all new to the department, are now part of its daily operations.

Underneath all of it is many years of deep, hands-on knowledge of computer hardware. It lets me get far more out of the computers and cloud services an organization already pays for, and avoid paying for capacity it does not need. I apply the same thinking to where software runs: across Azure, Cloudflare's global network and servers I run myself, I design systems that cost little to operate, stay reachable, and keep private services private. None of it came from standing still: two years of college, then twenty-five years of continuous learning, with every new technology learned by putting it to work.

Compute Efficiency: Getting More From Existing Hardware and Cloud Spend

Many organizations fix slow software by buying more: larger cloud servers, premium services, newer hardware. Often the real cause is somewhere else, and the extra spending does not solve it. I understand how a computer actually works, from its processors, memory, graphics card and storage to the connections that move data between them, such as PCI Express, USB and the network. I use that to find what is truly holding a system back before any money is spent. A typical modern laptop has far more computing power than most people realize. Knowing where its real limits are tells you whether a job needs a bigger cloud server, a different kind of server, hardware already on hand, or simply better-written software.

Diagnose before spending: Proved that a slow-loading AI model was limited by its storage drive, not by the link to the graphics card. That meant new memory or hardware would have saved at most four seconds. The effort went instead to the real one-minute delay, which no hardware purchase would have fixed.
Send each task to the right machine: Built a system that decides, for every request, which computer in a group should handle it. Demanding work goes to the most capable machine, and simple work to the quickest one that is good enough. Keeping the speech model loaded on the machine that runs it made transcription about five times faster (1.9 seconds instead of 9.3). Preferring the local network over the remote link cut network delay more than 100-fold (4.5 milliseconds instead of 574).
Share one graphics card between several AI tools: Designed a scheduler that lets language, image and speech AI share a single graphics card, and blocks any request that would overload it. An overloaded card gives no error message. It simply makes responses three to four times slower, which is why the scheduler prevents the overload rather than reporting it.
Match resources to the job: Sets each AI model to use only the memory and precision it needs. When a model sits idle, the hardware is freed automatically, so the same machine stays available for other work.
Use the whole machine: Spread demanding work across every processor core and the graphics card so a 3D game engine runs smoothly. Running a scan of four drives in parallel cut it from hours to minutes.
Run many machines as one pool: Manages seven Windows, Linux and handheld machines from one desktop as a single pool. Together they hold 108 processor threads, 277.2 GB of memory, 45.9 GB of graphics memory across three graphics cards, and 74.5 TB of storage. Fast solid-state drives hold working data, and large hard drives hold archives and backups. Work goes to whichever machine has an idle core and spare memory, up to 57 jobs at once over a 2.5-gigabit link to the main machine. AI models are parked to free memory before heavy jobs. Every machine reports its load and temperatures every 30 seconds, so a hardware purchase is decided from data.
Test on the hardware people really use: Measured performance on a high-end desktop graphics card, a mini PC's built-in graphics and a laptop. That way the results reflect real users' machines rather than one fast development computer.

Cloud, Edge and Network Architecture: Low Running Cost, Wide Reach

Using the cloud well is less about which services an organization buys than about how they are combined. I design where each part of a system runs: on a global edge network, in a cloud database, on a private server or on hardware already owned. The aim is a system that stays fast and reachable, costs as little as possible to operate, and never exposes anything private to the internet.

A complete game with no server of its own: Ship Voyage runs entirely on Cloudflare's network, so there is no game server to keep running. The edge serves its pages and game logic, saves sync to a cloud database (D1), and more than 41,000 voice clips stream from cloud storage (R2) with long-lived caching. All of its AI-written dialogue and voice was generated once, before release, so a new player adds no AI cost. If the cloud database is unavailable, the game carries on with the saves kept on the player's own device.
Public website, private back end: Built and deployed a multi-page business site on Cloudflare Pages whose chatbot and contact form reach a service on my own hardware without exposing it. A Cloudflare Worker passes those requests through an encrypted tunnel, so the back end never accepts connections from the internet and none of its files are published. A local test server reproduces Cloudflare's own behaviour, so problems surface before going live rather than after.
Cloud deployment made routine: Built a desktop tool that turns several command-line tools and a list of remembered steps into one window. It creates and redeploys Cloudflare Pages sites, scaffolds and deploys Workers with their routes, and creates Cloudflare Tunnels with their DNS records. It also installs the tunnel as a background service on Windows or Linux.
Private network across several machines: Connected a group of computers over WireGuard (an encrypted private network) and the local network. The system can tell a machine that is offline from one whose private link has merely dropped, and it plans the reconnection to match. It prefers the local network whenever it is available, and it moves files between machines at 112 MB/s with integrity checks, against 0.6 MB/s through an ordinary network share.
Off-site backups to a self-hosted server: Built and ran a console that copied every project's complete history to a self-hosted server on a schedule, over SSH and WireGuard. It confirmed that each copy matched and never overwrote history. When the private link was down, it skipped the run cleanly.
Players connected directly: Two of my games link players' computers directly to each other (WebRTC). A small Cloudflare Durable Object only introduces them, so there is no game server to keep running or pay for.

Selected results

About 2,000 a dayCompensation advisors using the pay and career history analysis tool I drove from idea to adoption with its lead developer.

Experience

Application Developer, HR and Pay AI Program

Feb 2024 – Present

Public Services and Procurement Canada (PSPC), Miramichi, NB

Led the hands-on technical work, in close collaboration with a fellow developer, subject-matter experts and the business, to conceive, engineer and deliver the MVP case-processing tools within the HR branch's AI case-processing project: membership dues, commuting allowance, leave with income averaging and executive performance pay.

No transcription, fewer errors: Aligned the tools' outputs with the Pay Centre's bulk-processing team so transactions, case notes and payment control were entered programmatically rather than retyped.
End-to-end data pipeline: Set up, with the team, the end-to-end process that kept payroll and case-management data current in an Azure SQL database: Azure Data Factory (ADF) pipelines and Oracle GoldenGate database connectivity, with concurrency and error handling, alongside SQL-based data access and control logic for the MVP prototypes.
more
Also designed the DuckDB and SQL schemas the pilot tools ran on, and ran the Protected B Microsoft Azure virtual machines that hosted the pilot.
Technical governance: Provided technical governance for a contracted AI document-processing program: reviewed and signed off on the contractor's technical approach and documentation, flagged gaps, and advised management on whether it was meeting its goals.
Image extraction: Built prototypes that extract case data from document images with OCR and AI, and collaborated on the data integration moving case-management image data into a departmental data repository.
Reusable foundation for production: Delivered the MVP's patterns, logic and business rules as a reusable foundation for the production version, with the requirements, technical documentation, code repositories, training material and knowledge transfer on how the ADF pipelines and databases connect, and coached colleagues in SQL optimization, Python integration and data architecture.

Application Developer, Pay Solutions Branch

Feb 2019 – Feb 2024

Public Services and Procurement Canada (PSPC), Miramichi, NB

Informal technical lead of the four-developer team that conceived, built and ran the Pay Centre's analysis and automation tools from 2016: holding the team together, setting direction and carrying each product from idea to adoption.

Pay History Analysis Tool: Drove the pay history analysis tool from idea to adoption: an Excel and VBA analytics tool built on data integration, joining 14 SQL queries of pay data with case details from the case management system into a day-by-day timeline of an employee's pay and career history.
more
It shows one cell for every day across each pay period and a bar for every payment, deduction, retroactive adjustment and job change. Every entry, cheque, earning and deduction opens its details on hover or in a popup, with links into the pay and case systems, in English and French; about 2,000 compensation advisors use it every day. Gathered what users and the business needed, built its back-end queries and data sources, cleared other work so its lead developer could focus, trained its users, kept it working through the move from Internet Explorer to Edge, Chrome and Firefox, and built a Citrix edition for departments outside the Pay Centre.
Upgrades and web version: Upgraded the original Excel and VBA tool with new SQL queries written to business requirements and published through change requests, brought its latest version, 4.0, through build and rollout, and led the requirements for its web version (TypeScript, HTML and SQL), available alongside it to every Government of Canada department.
more
Worked with the department's innovation team and a contractor through sprint planning, backlog grooming and data sharing.
Client Information Centre (a compensation tool): Co-developed and championed, with its creator, the companion tool for case work: built-in standard operating procedures, automated letters and case notes, and links into the pay systems, used with the analysis tool as one system.
Tools other teams still run: Built audit and overpayment-analysis queries and bulk patterns (finding the right employee record, filling one-time payment files in bulk) that overpayment, verification and other teams still run, including for client departments.
Queries with the Digital Services Branch: Worked with the Digital Services Branch to develop, maintain and upgrade the queries behind the team's tools, and carried them through business requirements and change requests so they are available to everyone.

5 more from this role

AI exploration: Technical advisor to a departmental AI exploration project; wrote the core SQL that captured complex pay business rules and supplied the project's data.
First cloud steps: Explored and prototyped on Microsoft Azure for pay processing applications, ahead of the cloud work on an AI case-processing project.
Strike leave-without-pay automation: Led the strike leave-without-pay automation, removing pay holds automatically and providing bi-weekly reviews and reports to stakeholders.
Compensation advisor tools: Integrated the a revision tool and a letter generator with the pay history analysis tool and case information tool, and supported compensation tools used by more than 2,000 active daily users.
Bulk processing: Built utilities for bulk processing of transactions, cases and payments in the payroll and case-management systems, and supported Cognos 10/11 user-acceptance testing and data validation for senior management and directors.

Business Expertise / Data Integrity Analyst

Dec 2015 – Feb 2019

Public Services and Procurement Canada (PSPC), Miramichi, NB

Learned the system to find the leverage -- Mastered the payroll's data, queries, policies and collective agreements well enough to see where automation would pay off most, then built it.

Learned the system to find the leverage: Mastered the payroll's data, queries, policies and collective agreements well enough to see where automation would pay off most, then built it.
The foundation the automation was built on: Built the SQL extracts, data tagging and quality reviews that the team's later case-closure, pay-history analysis and bulk-processing tools were built on.
Bulk payment automation: Delivered a bulk payment automation that cut operational delays and became early proof that RPA could work in traditional pay processing.
Data quality and reporting: Ran data quality reviews, recommended data and system improvements, and produced queries, tools and reports for management from payroll and case-management data.
Procedures and guides: Wrote and maintained standard operating procedures and system help guides with stakeholders.

Compensation Advisor

Dec 2014 – Dec 2015

Public Services and Procurement Canada (PSPC), Miramichi, NB

Advised employees and managers on pay, benefits and insurance, applying federal and provincial legislation and collective agreements.

Compensation advice: Advised employees and managers on a wide range of compensation and insurance benefits.
Legislation and agreements: Researched and applied acts, regulations and collective agreements, keeping current with compensation directives and changes.
Data integrity: Maintained accurate records across the federal payroll system, the pay system, the case-management system and PeopleSoft.
Quality assurance: Completed peer verification to confirm accuracy and adherence to policy.

Distribution Clerk/Client Service Officer,

Aug 2010 – Dec 2014

Service Canada (CSD), Bathurst, NB

Handled protected records, management reporting and employer inquiries for Service Canada programs.

8 more from this role

Administrative Expertise: Performed a broad range of administrative and clerical duties, ensuring efficient office operations and the integrity of document management systems.
Data Management and Reporting: Took charge of distributing, maintaining, and safeguarding both physical and electronic filing of protected documents, in addition to creating reports and statistics for management, enhancing organizational efficiency.
Client Support and Legislative Application: Delivered exceptional client service by answering and assisting employers with technical inquiries about Service Canada programs, applying government legislation, acts, and regulations daily to ensure compliance and informed client interactions.
Perform various administrative, and clerical duties
Distribute, maintain, and keep physical and electronic filing of protected documents
Create Reports/Statistics for management/senior management
Answer and assist employers with technical inquiries regarding Service Canada programs
Apply Government legislation, acts, and regulation on a daily basis

Technician

Jan - Sept 2009

PC Medic, Amherst NS

Supervised daily activities of PC Medic Store

3 more from this role

Supervised daily activities of PC Medic Store
Tracked inventory and sales history
Performed maintenance and repair on computers and networks

Payroll Administrator

2009 – 2010

Reg Plumbing and Heating, Moncton NB

Perform scheduling and supervision of staff

2 more from this role

Perform scheduling and supervision of staff
Payroll, collections, accounting, and various other duties

Material Purchasing Officer / Technical Support

Feb - Aug 2008

Benjamin Heating, Springhill NS

Performed all aspects of material purchasing for the production of furnaces and boilers

2 more from this role

Performed all aspects of material purchasing for the production of furnaces and boilers
Used soft skills to find best prices and quantity breaks for the company

Administrative Assistant

2008 – 2009

Correctional Services of Canada, Springhill NS

Coordinated and distributed daily mail throughout institution

4 more from this role

Coordinated and distributed daily mail throughout institution
Administered testing and interviews for positions within the institution
Completed inventory system update
Supervised in the distribution of inmate canteen

Technical Support

2007 – 2008

Rogers Cable, Moncton NB

Provided technical support for customers with internet, wireless data and cable issues

2 more from this role

Provided technical support for customers with internet, wireless data and cable issues
Met and exceeded quality and call handle times on a regular basis

Customer Service Representative

2003 – 2005

Telus Mobility, Barrie ON

Assisted clients with a multitude of cell phone billing, sales, and hardware requests

2 more from this role

Assisted clients with a multitude of cell phone billing, sales, and hardware requests
Used an array of applications to perform troubleshooting, billing, and sales analysis

Data Pod Technician / Operations Support Specialist

2002-2003 and 2006-2007

TeleTech Canada, Orillia, ON and Amherst NS

Assisted business clients with troubleshooting their wireless internet services

4 more from this role

Assisted business clients with troubleshooting their wireless internet services
Worked in a fast-paced call centre environment helping individuals and businesses with their digital communication needs
Coordinated with management and supervisors to ensure our clients' service levels were met
Used numerous applications to forecast call volume, create/maintain schedules, and produce detailed reports for the corporate management

Selected projects

AI Case-Processing Automation and Data Integration 2024–Present
Led the hands-on technical work, with a fellow developer and the business, on MVP automation for the department's AI case-processing initiative for pay and HR.
DevOps and Environment Administration 2022–Present
Run the Protected B Azure virtual machines that host the team's automation and AI applications in production.
Data Migration and Schema Modernization 2023–Present
I also designed the schemas and ETL pipelines (DuckDB, SQL) that the AI case-processing pilot and pay automation ran on.
Commuting Allowance and Leave with Income Averaging (LIA) Automation 2023–Present
Automated commuting-allowance and Leave with Income Averaging pay entitlements, removing manual case work equal to eight full-time positions.
Membership Dues and Auto-Closure Automation 2023–Present
Rules validate each case and close it automatically, and reports feed both audit tracking and management dashboards.
AI and Machine Learning Initiatives 2020–Present
Built the data foundation for the department's AI and machine-learning projects.
Pay History Analysis Tool Web Transition 2021–Present
Helped bring the pay history analysis tool to the web as a full web application available alongside the Excel tool.
Payroll Processing Innovations 2020–2023
Parameterized SQL and custom scripting did the calculations, with full audit traceability, so each result could be traced back to the rule that produced it.

Independent architecture and engineering

Project Cockpit: Control Center for AI Assistants 2026-05
Began as a control centre for running dozens of AI coding assistants at once, and became one AI service across seven machines. Applications use a single endpoint for chat, vision, speech and images. Each request goes to whichever machine is fastest or already has the model loaded, and the graphics cards are scheduled so that no model slows the others. Splitting work is where this pays off: a ten-minute recording divided across six machines finishes transcribing in 18.8 s. One machine alone took 138 s and silently returned 14% of the text. project details
CodeAtlas: Catalogue of Everything I Have Built 2026-08
Catalogues every software project I have built across four drives, so that I and the AI assistants I work with can find and reuse existing work instead of rebuilding it. It updates itself every night, and more than 1,500 automated tests check it first. That checking is the point, because a simple count of code is easy to get wrong and hard to notice. It also matches every Python environment on the machine against published vulnerabilities, and ranks what to fix by what is actually running. And it builds this resume, so every figure on it traces back to its evidence. project details
Ship Voyage (published as Ports of Call): Voiced Adventure Game Hosted Entirely on Cloudflare 2026-06
A voiced Age-of-Exploration adventure game that runs entirely on Cloudflare, with no server of its own. Its more than 41,000 voice clips were all generated before release, so a new player adds no AI cost. The clips stream from cloud storage, saves sync to a cloud database, and players can hold accounts and compete on a leaderboard. ship-voyage-4zn.pages.dev | project details
Ironvale: Real-Time Strategy Game With No Server 2026-09
A complete strategy game delivered as a single file, in which two players connect directly with no server in between. This works because both computers run the identical simulation, step for step. Since each side can compute the game for itself, only the players' commands need to cross the connection. ironvale.pages.dev | project details
Voxel Racers: A Python Racing Game Carried to the Browser, Proved Identical 2026-10
An arcade racing game built inside my voxel world, where pressing a key builds a course onto the land around you, then carried into the browser. A full rewrite would be slow and hard to prove faithful, so I split the work in two. Everything that does not change was baked in Python, and only the rules were rewritten in JavaScript. The browser version was then checked against recordings of the original: all 7,200 recorded physics ticks matched, and an eight-car race stayed within a millionth over 5,400. The baking runs across my fleet of machines, and the game plays on the web. voxel-racers-notes.pages.dev | project details
VoxelBlox: Open-World Building Game 2026-09
An open-world game for digging, building, driving and flying, in which code generates every texture, model, sound and piece of music. Its racing game reached the browser by baking everything in Python and rewriting only the rules in JavaScript. Recordings of the original then proved the port: all 7,200 recorded physics ticks matched. Multiplayer runs peer to peer with a signed handshake, and two machines measured 13-18 ms between them. voxelblox-notes.pages.dev | project details
VoxelStrike: 3D Combat Engine in Python 2026-05
A land, air and sea combat game over destructible terrain, written in Python. Python is not usually fast enough for games, so the slowest code paths are compiled and the terrain is drawn on the graphics card. Every automated check in it has been shown to fail when the thing it guards is deliberately broken, which proves each one can catch the failure it exists for. Its architecture became the starting kit for VoxelBlox. icemunk.github.io/voxelstrike-site | project details
NBLidar: New Brunswick as a 3D Block World 2026-10
Turns New Brunswick's open LiDAR surveys (laser measurements of the land) into a 1-metre block world anyone can explore in a browser, in English and French. Each square kilometre is built as its own job across a fleet of machines, so the source data never has to land on one computer. Nine Cloudflare Pages sites serve the result, which covers 5,139 km² so far. It reuses the mesher (the code that turns blocks into surfaces a graphics card can draw) and the browser-export method first built for VoxelBlox. nblidar.pages.dev | project details
Whisper Portal: Transcription and AI Code Assistant 2026-02
Turns large volumes of audio into text and identifies who is speaking, and grew into an AI assistant that investigates and edits code on its own. It is built to share the machine: it releases the graphics card when idle, and it sends work to other computers when they are better placed to do it. project details
Estate-wide security patching, proved before it ships 2026-10
Ran a security pass across every Python environment on my machines, testing each fix beside the old environment before it went in. Upgrading blindly breaks working applications, and not upgrading leaves known holes open, so every change had to prove it altered nothing else. CodeAtlas matched each installed package against published advisories, CISA's list of vulnerabilities exploited in the wild, and severity scores, then ranked the results by what was running. The runbook records 57 verified fixes. Six environments moved to a newer Python so their fixes could go in, and every fix left open has its reason written down. The sweep now runs every night; 177 vulnerable installs remain in live environments, and ten of them are on CISA's list. project details
AerobizX: Airline Strategy Simulation 2025-12
A complete airline-management strategy game whose whole economic simulation runs in the player's web browser. The player builds route networks, weighs the economics of a fleet and competes for passenger demand against rival airlines that react to each move. Because the browser does all of the work, there is no server to run. project details
Energy Equity Research Platform 2026-03
A research dashboard for energy-sector investments, built so that every figure can be traced back to how it was calculated. AI analyses stay on file as records rather than being thrown away, so a conclusion can be revisited later. Any earlier state of the research can be restored exactly. project details
Equity Screener and Research Desk 2026-01
A stock-research tool that gathers market data with an automated browser and refreshes only what is missing or out of date. It stores each research experiment together with the data it used. That makes a past screen repeatable: it can be replayed against the data as it was at the time, not as it is today. project details
WhisperKeyboard (published as Ownword): Voice-First Android Keyboard 2026-08
An Android keyboard built around speech, with transcription on my own server and prediction and autocorrect on the device. When the original source code was lost, I rebuilt it from the released app. I then checked the reconstruction against the original, step by step. keyboard.rnsu.net | project details
Document Pipeline Orchestrator 2026-03
One web-based system for processing documents, replacing two older desktop applications. It checks the result of every step before moving on. When a step fails, the job resumes from that point rather than starting over, so a late failure no longer costs the work that came before it. project details
D-Docs: Document Publishing System 2026-01
A document publishing system that uses AI to produce text, images and speech. Its most useful lesson came from a performance study. Measuring first showed where the time was really being lost, and that kept the effort away from a fix that would not have helped. project details
Miramichi Intelligent Automations: Business Website 2025-09
The website for my automation and applied-AI business, designed so the public pages and the private back end never meet directly. The pages are static and served from Cloudflare's network. A Cloudflare Worker passes chatbot and contact requests through an encrypted tunnel to a back end on my own hardware. A local test server behaves the way Cloudflare does, so problems show up before the site goes live rather than after. www.rnsu.net | project details
Pay history analysis tool modernization: VBA to Python (prototype) 2026-03
In the prototype those five became one.

Writing

Engineering a Voxel World 2026-09
An architecture and engineering case study of an endless, procedurally generated voxel world in Python: the decisions and trade-offs, how it was verified, incident reviews, and a graphics-card renderer built to meet its frame budget. voxelblox-notes.pages.dev
VoxelStrike: a Combat Game on a 1992 Terrain Trick 2026-09
A plain-language account of how VoxelStrike is built: a 1992 landscape-drawing technique rebuilt in Python and moved to the graphics card, an endless world, vehicles made of boxes, and how every problem was measured and fixed. voxelstrike-notes.pages.dev

Skills

Programming Languages and Scripting SQL for complex queries and ETL, Python for automation, AI integration, and data transformation, VBA for Office automation, PowerShell for Windows and system automation, and C#/VB for application development. Strong fundamentals and rapid adaptation to new frameworks and languages.
Automation and Orchestration Designed and deployed multi-stage pipelines combining OCR, structured JSON generation, validation, and calculator loops to automate complex HR and pay cases. Scheduled and idempotent jobs with full logging, retries, and audit trails.
Cloud Computing and Virtualization Protected B Azure environments, VM administration, secure storage, API integrations, reverse proxies, and deployment automation. Managed production applications and services for multiple user groups.
Data Engineering and Migration End-to-end schema design using DuckDB and SQL; Python-driven extraction and transformation; large-scale migration of 500 000+ protected images and datasets to cloud storage with structured, auditable data pipelines; Azure Data Factory (ADF) pipelines and Oracle GoldenGate database connectivity, with queries tested against the connected data.
AI and Intelligent Document Processing Implemented Azure Document Intelligence for OCR and layout analysis; integrated OpenAI-compatible LLMs for JSON transformation, validation, and deterministic rule enforcement in pay and HR automations.
Business Intelligence Tools Advanced Cognos and Power BI development for interactive dashboards and automated reporting, supporting both operational and executive decision-making.
Robotic Process Automation (RPA) Automated the federal payroll system, the case-management system and PeopleSoft operations to remove manual steps and reduce errors, executing tens of thousands of validated transactions.
Database Management and Analytics Managed large datasets, performed cleansing and analysis using SQL, Python, C#, and VBA. Built robust ETL pipelines ensuring data integrity and traceability.
High-Volume SQL Delivery Authored 2300 + ad-hoc SQL extracts for Pay Centre managers and analysts, dramatically improving responsiveness and throughput.
Software Development Methodologies Agile and full SDLC execution from inception to maintenance; API-first design, configuration via .env and JSON settings, and audit-ready release controls.
Web Development Built Python-based web applications and services using HTML, CSS, and JavaScript to support user-facing automation workflows focused on accessibility and intuitive design.
Systems Integration Engineered the federal payroll system and PeopleSoft data synchronization and automated validation routines that maintain consistency and prevent duplication across systems.
Security and Compliance Applied Protected B safeguards, encrypted storage and transfer, strict retention policies, access control, and comprehensive logging for audit readiness.
Leadership, Communication, and Knowledge Transfer Coordinated developers, SMEs, contractors, and business stakeholders. Produced diagrams, documentation, guides, and video walkthroughs to enable users and sustain adoption.
Project Management and Delivery Scoped, planned, and executed AI, automation, and data initiatives on time and within budget, emphasizing measurable efficiency and operational value.
AI-Assisted and Agentic Engineering Building quickly with AI coding assistants while keeping the result dependable and auditable. Agentic development, specification-driven design, automated regression testing, resumable and idempotent pipelines, Model Context Protocol (MCP) integrations, audit trails.
Cloud, Edge and Network Architecture Designing where software runs so it costs little and stays reachable. Cloudflare Pages, Workers, Pages Functions, D1, R2, Durable Objects and Tunnels, DNS; GitHub Pages; Microsoft Azure virtual machines; WireGuard and SSH private networking; reverse proxies; Windows and Linux background services; scheduled, verified backups; deployment automation.
Hardware, Performance and Compute Efficiency Getting the most from existing hardware and cloud resources, and sizing them to the work. CPU, memory, GPU and storage performance, PCI Express, USB and network bandwidth, GPU sharing across AI models, model quantization, parallel processing, NVMe over TCP, performance profiling, capacity planning and cost optimization.

Competencies

Client Service Results 3
- Specialized Technical Advice and Support Advise across the full software development life cycle, shaping tools like the pay history analysis tool and case information tool around how compensation advisors actually work.
- Developing and Supporting Work Processes and Standards Wrote the technical reports behind the strategy for moving the analysis tool to the web, and built an ongoing acting-pay automation script that standardized acting-case closure.
- Advisory Role in Software Solutions Policies and Standards Advise program and technical staff on software policies, standards and federal guidelines, including for AI data work.
Collaborate with Partners and Stakeholders 2
- Building and Nurturing Effective Networks Aligned internal teams, the department's innovation team and IBM contractors to deliver the web version of the analysis tool to its functional and performance standards.
- Active Listening and Understanding Stakeholder Perspectives Gathered needs from data scientists, compensation advisors and technical experts to shape AI data extracts that are both innovative and practical.
Create Vision and Strategy 3
- Visionary Leadership in Compensation Management Set the technical direction for pay data integrity from the team's founding in 2016, building the SQL extracts and tools that compensation advisors rely on, a pay history analysis tool and a case information tool among them.
- Strategic Integration of Advanced Technologies Brought machine learning, OCR and cloud into compensation work: built core data extracts for AI projects, provisioned secure Microsoft Azure environments, and wrote REST APIs linking the payroll system to cloud databases.
- Engaging Stakeholders and Building Consensus Built support for moving the analysis tool to the web by working with the compensation advisor community, gathering feedback and refining the pay history analysis tool and case information tool until they became part of daily work.
Mobilize People 4
- Inspiring High-Performing Teams Led the modernization of an acting-pay automation tool, applying mathematics and logic to review entire acting accounts, calculate daily pay and flag discrepancies, a tool another team still runs today.
- Fostering a Collaborative and Inclusive Work Environment Build tools with the people who use them: the pay history analysis tool and case information tool were shaped by compensation advisors' input, and a departmental AI project brought data, pay and technical experts into one team.
- Commitment to Continuous Learning and Development Completed advanced training in AI, machine learning and cloud technology, applied it to AI data extracts and Azure research, and pass that knowledge on to the team.
- Providing Constructive Feedback and Managing Non-Performance Led analysis-tool query modernization and case information tool testing with clear, measurable goals and direct feedback, delivering high-quality results on time.
Promote Innovation and Guide Change 3
- Championing Innovation and Embracing Change Introduced machine learning, OCR and advanced analytics to the Pay Centre through a departmental AI project and core AI data extracts.
- Encouraging Creative Approaches and Risk Management Guided Python and Power BI scripts that find missing pay steps, balancing new approaches with sound risk management.
- Learning from Setbacks and Driving Continuous Improvement Refined retroactive pay and maternity top-up automations from real-world feedback until they exceeded their objectives.
Technical Expertise 3
- Deep Knowledge of IT Theories, Principles, and Practices Apply IT architecture, machine learning, OCR and secure Azure provisioning to improve accuracy and efficiency at the Pay Centre.
- Communication and Negotiation Skills Explain technical requirements to contractors, suppliers and staff through research papers, specifications and technical reports.
- Mentorship and Knowledge Sharing Mentor junior staff in AI, machine learning and software practice through training sessions, workshops and one-on-one coaching.
Uphold Integrity and Respect 2
- Creating an Inclusive and Respectful Work Environment Ran the strike leave-without-pay automation as a shared effort, involving the whole team in decisions on complex hold-removal logic.
- Supporting Workplace Well-Being Keep workloads balanced and teams healthy through demanding projects, drawing on mental-health training.

Education and continuous learning

DeVry Institute of Technology: Computer Information Systems (two years of study, the foundation everything since was built on)
Continuous learning: Twenty-five years of steady upskilling since college, from technical support to leading AI development, learning each new technology by putting it into production rather than only studying it.
Professional development: Continual, hands-on upgrading across AI and machine learning, cloud platforms, business intelligence and reporting, robotic process automation, data engineering and software development, through training programs, research and development work, and applied projects rather than a single credential.
Self-directed engineering: Learned Python, JavaScript, TypeScript, Kotlin, GPU computing and modern AI tooling after college by building more than 90 independent software projects.
Teaching others: Trained colleagues on the compensation tools I built and coach the team in SQL, Python and data architecture.
Advanced training in AI, machine learning and cloud technology: Formal training in artificial intelligence, machine learning and cloud platforms. Applied to: core data extracts for AI projects and the department's Azure research and development.
Microsoft Azure research and development: Hands-on research into Microsoft Azure services, carried from evaluation through to production. Applied to: secure cloud environments and AI services now in daily use.
Robotic process automation (RPA): Robotic process automation training with the department's innovation team. Applied to: rule-driven automation of high-volume pay cases.
Business intelligence and reporting (Cognos, Power BI): Business intelligence and reporting tools for operational dashboards and analytics. Applied to: Pay Centre dashboards and anomaly detection.
Workplace mental-health training: Training in recognising and responding to workplace well-being concerns. Applied to: keeping workloads balanced and teams healthy through demanding projects.