Richard Amerman
Building systems, exploring the foundations of machine intelligence, and questioning what genuinely capable AI might require
Current Focus
My primary intellectual energy is split between two directions. The first is independent exploration of what genuinely capable AI systems would actually need to look like. Not incremental improvements to current LLMs, but fundamental questions about comprehension, continuous learning, and how machines might build dimensional models of the world they operate in. I published the first article introducing my knowledge distillation framework, which lays out an understanding-based approach to machine intelligence. The second is building: a voice AI learning platform, a codebase comprehension tool that refuses to claim more than it can prove, a country comparison platform for people planning a move abroad, and the multi-agent ecosystem I run all of it from.
I respect the fact that I don't have the formal credentials to claim authority in this space, and I'm only beginning to engage deeply with the broader literature. My professional engineering work in cloud infrastructure and AI-assisted development provides the practical foundation for everything I build. I'm also active in the Portland and Vancouver AI community, presenting regularly on voice AI, multi-agent systems, and architecture comprehension, and increasingly sought out as a mentor and voice AI resource. You can see where I've spoken.
Projects
A mix of open source, source available, and private work. Every one says plainly which it is. See all of them, with the full story on each.
SysCorpus
Complete, navigable maps of real codebases, built on the rule that the map never says more than it can prove. Every file is accounted for as parsed or skipped with a stated reason, and when the AI layer cannot prove an answer it files the gap instead of guessing. Maps VS Code, roughly 5 million lines, in about three minutes.
Source available
UnaMentis
Open-source voice AI learning platform, and the origin project that started this journey. Provider-agnostic architecture lets you swap STT, TTS, and LLM freely. iOS and web. Learning through natural conversation, anywhere.
Open source
The Agent Ecosystem
Four devices running as one system so AI agents can do real work here all day, with a human able to see and steer it. A findings queue, a dashboard, a physical control deck I built on hardware whose USB protocol I had to work out myself, and a watch app for the glance that answers whether anything needs me.
Private
Expat Intel
Country comparison for people planning a move abroad. Over 50 metrics across visas, cost of living, taxes, healthcare, safety, and climate, assembled from a dozen open data sources, with a public status page showing which of those sources are healthy right now.
Private
Pocket TTS for iOS
Ported Kyutai's Pocket TTS (117M parameter model) from Python/PyTorch to Rust/Candle for native on-device iOS inference. Solved 14 major technical challenges across 5 specialized AI agents, achieving near-identical waveform output to the original model.
Open source
OpenClaw Managed Cluster
A deployment package that turns OpenClaw from a powerful but unsafe-by-default local agent into something an operator can run safely at scale. Infrastructure as code, a management web app, an observability stack, zero credentials on the agent VMs, and private inference as a first-class target.
Private
Core Expertise
AI Systems Thinking
Independent exploration of model architecture, safety constraints, and the structural requirements for systems with genuine comprehension. Early-stage thinking on symbolic-neural integration, continuous learning, and representational grounding. I can't help but question our continued use of very old design and representation of neurons in technology.
Applied AI Engineering
Daily building with Claude Code and multi-agent systems. Experience designing governed agentic workflows, voice AI architectures, and AI-assisted development patterns that actually ship to production.
Software Engineering & Systems
30+ years across development, analysis, architectural design, and infrastructure. The first half of my career was predominantly software development (Python, C#, Java), and the second half has been cloud infrastructure and DevOps (Azure, Terraform, Kubernetes, CI/CD). Full stack from bare metal to distributed services.
Core Experiences
The Influence of an Island Town in Alaska
I grew up in Sitka, AK, a small town in SE Alaska. I can always look back on my time growing up there as a key foundation for the rest of my life. For me and many others in Sitka, our world was not this small town on an isolated island, but the entirety of the real world.
Public Radio: KCAW
Started working in public radio in my hometown of Sitka in 8th grade. Was trained, certified, and was on the air live with one or another show throughout high school. To this day, I do an air shift every time I'm back in my hometown.
Presidential Honor Guard
Joined the US Army and was recruited into the 3rd U.S. Infantry Regiment, Old Guard, while in basic training at Fort Benning, Georgia. I was then stationed at Fort Myer, Virginia, the home of The Old Guard, the Presidential Honor Guard directly after basic training.
Old Guard Fife & Drum Corps
While stationed at Fort Myer in the Old Guard, I decided I needed a significant change. While my options were few, one of them, while ambitious, was the Old Guard Fife & Drum Corps. I was able to teach myself fife, pass the tryout, and spent the rest of my 4 years in the FDC.
Hang Gliding Instructor
I learned to hang glide at Kitty Hawk Kites in Nags Head, NC soon after I left the Army. After flying there and hanging out with the instructors for a few weeks, I was offered a job as an instructor. I came back a few months before the next summer season, and spent every day learning until I was ready to teach. I taught there for 2 years full time.
EMT and First Aid/CPR Instructor
While I was teaching hang gliding, I first became a first responder as a requirement of the hang gliding instructor job. I then went on to become an EMT-B, and certified First Aid/CPR instructor for both the Red Cross and the American Heart Association.
Traditional Irish Musician
I started playing the Irish tin whistle when I was in the Fife and Drum Corps. Over the years I added Irish flute as my primary instrument and have been playing it ever since. The highlight of my Irish music career was during the '90s in Seattle when I was a member of multiple active gigging bands and also active in the Irish step and set dancing community.
Accepting a Tech Career
My plan from high school on had been to become a music teacher, and I held onto that plan well into my early 30s. I finally gave in to a career in tech that had been developing since high school.
FIRST Robotics Mentor
Last year I became a mentor on our local FIRST Robotics high school team. It was definitely one of the greatest experiences I've had in my life. It was incredible to see the effect this program had on the high school kids on this team.
Exploring the Foundations of Machine Intelligence
In 2025, after decades in infrastructure and development, I dove into the deep end. Not just using the tools, but questioning what these systems actually are, what they lack, and what genuine machine intelligence might require. This led to publishing a knowledge distillation framework, porting a 117M parameter TTS model to iOS, building governed multi-agent systems, and becoming an active voice in Portland's AI community. This has become the most intellectually alive period of my career.
Professional Journey
Independent AI Exploration & Development
2026
Published the knowledge distillation framework on Substack, whose working subtitle is knowledge distillation into a dimensional world model. Shipped the Pocket TTS iOS port (Rust/Candle) and the OpenClaw Managed Cluster deployment package. Took Solution Explorer to SysCorpus, adding the evidence and adjudication layers that let a generated map be trusted. Launched Expat Intel. Built the four-device agent ecosystem, including a physical control deck and a watch app, and continued development on UnaMentis. Presenting regularly in the Portland and Vancouver AI community.
Independent AI Exploration & Development
2025
Began building open-source AI projects (UnaMentis, Agent Vision Team, Solution Explorer), exploring model architecture and AI safety, and writing on AI systems design.
CD Baby (Downtown Music Holdings) - Senior Cloud Engineer
2014 - Present
Eleven-plus years and two acquisitions, from CD Baby to Downtown Music Holdings to Virgin Music Group. A 700TB+ enterprise SAN, fully virtualized infrastructure, Azure with Terraform and Kubernetes, a GitHub Actions to ArgoCD GitOps migration, and an audio ingestion system for digitizing artists' CDs. From 2024 onward, systematically bringing LLMs into our development workflows and becoming the go-to AI resource in the organization.
Diverse Technology Leadership
1994 - 2014
20 years across multiple industries: RBA International, Columbia Analytical Services, PEMCO, and founded 7 Tech Northwest.
Unique Beginnings
1987 - 1993
From the US Army Old Guard Fife and Drum Corps to hang gliding instruction, building the pattern of embracing radically new challenges.
Two Ways to Go Deeper
Ask about me
There's an AI assistant on this site that knows my work, history, and thinking in depth, not just the resume facts. Ask it why I'm making this change, what I actually built, or what I think today's models are missing.
The candid version
I wrote a long FAQ answering the questions I'd want answered if I were reading this. Why the career change this late, what I think I actually offer, and where I'm honest about the gaps.
"I don't yet know what I don't know about these systems. But I know how to find out, and I know it matters more than anything I've worked on before."