Matthew T. Kemp III

Stim Technician · Chemical Technician

Patterson-UTI / Started January 2022

• Used real-time data acquisition software to manage chemical inventories and coordinate chemical orders and deliveries.
• Compared actual and metered volumes, maintained chemical and sand QC records, and recorded job documentation.
• Troubleshot serial, TCP/IP and cloud connections; recorded wellhead and operational status information.
• Coordinated across teams and customers to support project execution.

Equipment Operator III · Supervisor in Training

Patterson-UTI · Universal Pressure Pumping / October 2016 – December 2019

• Provided supervision and leadership to a crew of 15 and supported employee development.
• Created price tickets per well stage, including chemicals, sand and run time.
• Worked with zipper frac and single-well operations, pump-room and blender operation, Excel, Dastrac and FracPro.
• Led JSA and safety meetings.
• Recognized by the company for job-site awareness: while operating the Solaris sand management system, he noticed a co-worker enter a silica-regulated area without respiratory protection, removed the person from the area, and explained that entry requires a respirator.

Field · Pumping Specialist

Pioneer Natural Resources / July 2013 – January 2015

• Calculated cement volumes and acid jobs and coordinated job design and scheduling with drilling teams.
• Trained and assessed three to five equipment operators during three-month training periods.
• Worked across coil tubing, fracturing, cementing, acidizing and drilling operations.

District Safety Mentor · Frac & Cement Operator

Baker Hughes · BJ Services / July 2010 – May 2013

• Trained 15–20 employees in operational, hazard and safety requirements and mentored employees during training schedules.
• Worked with chemical-additive, blender, hydration, cement, bulk-truck, nitrogen and related field equipment.
• Collaborated across departments to support field operations.

ShiftLedger · AI-assisted workflow prototype

• Specified and directed a materials-reconciliation, follow-up and shift-handover application from field workflow knowledge, with calculations running on a server-side JavaScript Worker.
• Defined the rules that keep records trustworthy: unknown values stay unknown, signed shortages and surpluses stay separate, costs that are missing stay separate from known costs, and a review must name a person.
• Verified with 61 automated tests covering calculations, access and origin controls, source exposure, workflow state and stale responses. Confirmed document and data downloads in a real browser.

Grounded assistant evaluation study · AI Lab

• Designed and tested a permission-aware, source-grounded handover assistant pipeline: authorize, retrieve, draft, verify, with a model-free baseline and a slot where a language model could be added.
• Wrote 16 evaluation cases across answerable, outdated, unauthorized, conflicting, prompt-injection and unanswerable questions. 15 required cases pass, and one known limitation is published.
• Added 9 tests, including a property test that checks every reader against every record for leakage, and mutation checks showing the tests fail when a safeguard is removed.
• Measured latency per question in the browser, with zero model calls and zero model cost, and recorded the tradeoffs in a decision log.

AI-assisted build process

• Directed Claude and Claude Code from written project instructions and a running state file, with tests written before features, deliberate mutation checks, and browser verification at desktop and phone widths under a strict content-security policy.
• Kept humans in charge of domain rules and public wording; documented what is not built. Details on AI & systems.

Field Notes content workflow

• Configured a recurring drafting workflow that prepares LinkedIn post drafts from approved project evidence, checks claims, keeps employer information private, and requires human review before anything is published. It does not publish directly.

Bachelor of Business Administration in Finance

Sam Houston State University, Huntsville / December 2009

ShiftLedger

AI-assisted evaluation prototype / Matthew Kemp

A working, field-informed prototype for materials reconciliation, accountable follow-ups, shift handover and management visibility. Intended for company evaluation, potential licensing, and tailored workflow projects. Uses fictional data, HTML, CSS, JavaScript and server-side calculations. Development evidence and limitations are documented in the project case study.