Ian MackinnonAbout · cross-section

About

My career drawn as an SEM cross-section, a nod to my years as my research group's nanofab super user. Time runs bottom to top, one year to the scale bar, and work that overlapped sits side by side like grains.

Career cross-section

Choose a stretch of work to highlight
πr²x² + y²now · still growing×3×3×3×3×3×3×3200820132017202020232026
SE2 · cross-section 2008 to 2026 · IM1 year
1 · 2008 to 2013

Materials science research

Controlled experiments, calibration, and asking whether an effect is absent or my approach needs refinement.

×3 texture: electrospun fibers and nanoparticles
2 · 2013 to 2023

Education Made Visual (tutoring company)

Explaining in visual ways from first principles (300+ students, 5,000 hours of instruction).

×3 texture: a chalkboard of worked equations
3 · 2017 to 2019

SkyeLight Coffee Company (co-founder)

Taking a product from a recipe to small-batch production to local wholesale.

×3 texture: roasted beans
4 · 2019 to 2020

Triumph Group (composites R&D)

Helping develop a new welding process (the first fully welded test article with no separate heating layer) and writing test plans that set the work for technicians and machinists.

×3 texture: carbon-fiber cross-sections in a thermoplastic matrix
5 · 2020 to 2022

Airborne Imaging (aerospace programs)

Budgets, contracts, regulatory compliance, and winning proposals across 19 projects and 8 organizations.

×3 texture: aircraft laminate, with the occasional plane
6 · 2022 to now

Mission-driven health and wellness company

Launch deadlines for a large audience, then building internal and external AI systems (sole AI engineer since May 2025).

×3 texture: a network of connected systems
7 · 2026 to now

IRM Forge and independent research

Building my own tools and running my own research program.

×3 texture: fresh crystal facets, the newest growth

IN MORE WORDS

I started in materials science, with a B.S. (summa cum laude) from the University of North Texas and an M.Eng. from the University of Texas at Arlington. From 2008 to 2013 I worked as a research assistant on nanoscale materials, biosensors, and biomedical coatings, served as my group’s nanofab super user (SEM/EDS, XRD, confocal microscopy, Class-100 cleanroom), and built a 20 kV electrospinning system for collagen-fiber research.

At Triumph Group I worked on electromagnetic welding of thermoplastic carbon-fiber composites and helped produce the first fully welded test article with no separate heating layer. At Airborne Imaging I was the company’s first dedicated project manager, running a program of 19 aircraft-modification projects (over $2.5 million) across 8 organizations, including MIT Lincoln Laboratory.

From 2013 to 2023 I ran Education Made Visual, a tutoring company, and tutored more than 300 high school and college students in math and science (5,000 hours of instruction, with a 96% pass rate). I also co-founded SkyeLight Coffee Company, a health-focused, real-ingredient wholesale coffee company.

In 2022 I joined a mission-driven health and wellness company on the launch team, which hit every partner-coordinated launch date across 61 projects, and I’ve been promoted twice. I report to the COO and have been the company’s sole AI engineer since May 2025, with 11 AI builds shipped (including a production RAG chatbot and a launch-setup workflow that went from about 80 person-hours to about one over three iterations). I also lead the company’s evaluation of Gemini Enterprise agents for specific workflows, run staff trainings on working with AI, and gave input on its AI-use policy.

In 2026 I started IRM Forge, my independent studio for software products, where I built fleet-pm and Alloybase. Alongside that I run independent research on how language models work and how to evaluate them.

WHAT I'M WORKING TOWARD

I want AI to go well, and to be truly beneficial for everyone it touches: people, animals, the planet, and the generations after us.

I’m working toward a solarpunk future, where technology is sustainable, decentralized, legible, and scaled to the communities that use it (an idea I took from Joshua Krook’s essay New Intrigue), and I think AI can help build it. The work to make that happen starts now, and I want to be an integral part of it.

That’s why I build AI a small organization can run and control itself (local, quantized models), why I study interpretability (systems people can see into), and why I’d like to put materials tools like Alloybase to work on energy-transition materials. The AI risk I think about most is power concentrating with whoever controls the most capable systems, because I think it feeds most of the other harms (disempowerment, cyber and biological risk, and wider economic gaps within and between countries).

The causes I care most about are animal welfare and the environment.

HOW I BUILD

I look for the workflows where AI saves people real time, build the tool with the people who will use it, and design it so they can trust what it gives them. In practice that means anything that has to be correct runs in code, the model handles the parts that need judgment, and every AI feature has a defined way to fail. I think a solid feedback and maintenance process does more to keep people using a tool than any single feature does.

IN NUMBERS

$2.5M+across 19 aircraft-modification projects, 2020 to 2022

$0$2.5M+program budgetprojects in order of start$

Each plane is one project, climbing in even steps to the program total. Flags and markers show counts, not which projects.

  • 9 came from proposals I wrote and won
  • 18 of 19 on original budget (94%)
  • 17 of 19 on original schedule (89%)
  • All 19 delivered to quality standards.
I was Airborne Imaging's first dedicated project manager, running 19 aircraft-modification projects across 8 organizations, including MIT Lincoln Laboratory.

96%pass rate, teaching math and science visually

A lesson in the style I taught: why a circle's area is πr²base = 2πrr

Ring 1 sits 4 units from the center, so it unrolls into a strip 2π × 4, about 25 units long. Ring 2 sits 12 units from the center, so it unrolls into a strip 2π × 12, about 75 units long. Ring 3 sits 20 units from the center, so it unrolls into a strip 2π × 20, about 126 units long. Ring 4 sits 28 units from the center, so it unrolls into a strip 2π × 28, about 176 units long. Ring 5 sits 36 units from the center, so it unrolls into a strip 2π × 36, about 226 units long. Ring 6 sits 44 units from the center, so it unrolls into a strip 2π × 44, about 276 units long. Ring 7 sits 52 units from the center, so it unrolls into a strip 2π × 52, about 327 units long. Ring 8 sits 60 units from the center, so it unrolls into a strip 2π × 60, about 377 units long. Stack every strip and they make a triangle: base 2πr, height r. Area = ½ × 2πr × r = πr².

a 96% pass rate, shown as 96 of 100

  • 300+ high school and college students
  • 5,000 hours of instruction
  • 10 years running Education Made Visual
From 2013 to 2023 I ran Education Made Visual, a tutoring company built on explaining math and science visually, from first principles (300+ high school and college students, 5,000 hours of instruction).

CERTIFICATIONS

  • PMP, Project Management Professional (2025)
  • ISO/IEC 42001 Internal Auditor (Exemplar Global TPECS, 2026)
  • BlueDot Impact: AGI Strategy and Future of AI (2026)
  • AIGP: in progress
  • CAPM (2022) and PCEP (2021)

OUTSIDE WORK · CONTACT

Outside work I read fiction, roast coffee, and explore the outdoors.

The best way to reach me is on LinkedIn.

Built with Astro and developed with Claude Code. All copy reviewed and approved by Ian.