Evidence brief · Augusto Yuri Villa · November 2026

Twenty years of machinery business, now building applied AI.

Five dimensions, each backed by numbers and a source you can check: the industry record from IPESA, Atlas Copco and Epiroc, the M.S. research at NC State, and what I built in 2026 at PneuraCore and on GitHub.

years in industry
20+
lines of code, 8 main repos
~122k
automated test files
238
GitHub repositories
14
live services
17
publications & theses
7

Technical build & ship

Designs and runs production AI systems end to end, with tests, audits and human approval gates, and keeps a mechanical-engineering base in CAD and Ansys.

  1. 360+ commitsPython and TypeScript across 14 private repos, Feb–Oct 2026; largest are SBTest (171 commits, 21.6k LOC) and the PneuraCore ecosystem (95 commits, 42.7k LOC).GitHub · yvillajd-netizen
  2. 4 pluginsPrivate Claude Code marketplace: paper-harness v1.10.2, visual-harness v1.3.1, sbtest-review, pneura-video-creator.pneuracore-plugins · Pneura_video_creator
  3. Edge + localSBTest review sandbox: the public side has no path back to the machine that holds the documents; hashed expiring reviewer tokens, agents that turn comments into approved edits.sbtest-pneuracore
  4. 17 servicesSelf-hosted ecosystem I designed and run, containerized, with Zero Trust login; 16 of 17 operational at the last check.PneuraCore service directory · verified 20 Aug 2026
  5. 6 scansClaude Security scans and four dated static and security audits on the PneuraCore monorepo before deployment.pneuracore_ecosystem
  6. FEA · IoTAnsys validation of an excavator adapter plate and a hammer cage; ESP32 plant-sensor node from wiring to enclosure.Ansys · ESP32 · NC State lab practice

Commercial revenue & market

Grew dealer networks, product lines and share in four machinery sectors, and now turns that market sense into lead, marketing and launch tools.

  1. $17M / yrBusiness model built at Atlas Copco, with a 10% gross-profit rise and $2M from portfolio optimization.Atlas Copco 2009–2018 · LinkedIn
  2. $5M · 50% GPPioneering machinery rental business, three years; dealer network expansion added another $5M a year.Atlas Copco · LinkedIn
  3. +40% revenueEpiroc key accounts in Mexican mining through contract stock optimization; $70M record annual customer revenue managed.Epiroc 2018–2020 · LinkedIn
  4. 5% → 40%Helped grow John Deere's share in Peru as product specialist and market analyst.IPESA 2003–2009 · LinkedIn
  5. 3 toolsGMaps lead generator, BIG4 weekly GA4 marketing intelligence with a four-agent chain, and the Sweet-APPS launch for packers.gmaps-lead-generator · big4-pneuracore

Managerial operate & govern

Runs operations with KPIs and cost control, and builds software the same way: one source of truth, dated audits, nothing irreversible without sign-off.

  1. $2M / yrSavings from the run-off and overhaul test center I proposed and launched.Epiroc Mexico 2019–2020
  2. 25+ dealersManaged across industry, construction and mining as member of the division management team.Atlas Copco Latin America
  3. Data → stateHICAMS-DOT ecosystem for Crumpler Plastic Pipe: ETL and structured project data delivered to a state transportation agency, 33 commits, 96 test files.hicams_CPP
  4. GatedHarness rule set: SHA-256-verified masters, evidence gate before any edit, and --approve only a human can give.Research_eva · paper-harness plugin

Leadership people & partners

Led large cross-functional groups and partner negotiations in Latin America, and teaches what I build so others can run it.

  1. 200+ · 300+Cross-functional team led at Atlas Copco; internal stakeholders coordinated at Epiroc.LinkedIn
  2. 2 countriesLed the corporate negotiation to become BROKK's demolition-robot dealer in Peru and Bolivia.Atlas Copco
  3. Team of 4Led design and fabrication at NC State as research assistant; teaching assistant; NC State Leadership Development Program.NC State 2021–2023
  4. ExplainersStep-by-step guides so operators and reviewers can use the systems: console walkthroughs, reviewer guides, an ESP32 primer.sbtest-pneuracore · paper-harness

Research method & evidence

Publishes machine-learning work on real machinery and applies the same discipline to auditing models and manuscripts line by line.

  1. R² 0.93Random forest predicts tractor noise and flags atypical sound levels; JETA (MDPI) 2026, doi:10.3390/jeta4040036.Publications · 7 incl. M.S. and B.S. theses
  2. 95%Accuracy of the R random-forest sales forecast for Northwest Mexico mining.Epiroc 2019
  3. 12 findingsForensic code-vs-manuscript audit of the DSSAT SUBSTOR-Sweetpotato model, one documented card per discrepancy plus an inventory and change dictionary.DSSAT SUBSTOR-Sweetpotato audit · 2026
  4. ASABE ×2Presented research at ASABE 2022 and 2023; ASABE certified reviewer; patent collaborator 2023–2024.NC State BAE