Principal AI Engineer & Architect
Aric Camarata
Aric served in the U.S. Army from 2001 to 2005, where he developed the discipline and composure under pressure that still defines how he works. He had been writing software since middle school.
He launched his development career at Roxbury Group in 2003 and spent the next decade shipping enterprise systems for Apple, AT&T, Capgemini, Pearson, and Northrop Grumman before founding Unity in 2014. The Unity decade covered enterprise contracting for Credit Karma, Walmart, Cigna, Cambium, Progress, and CMD plus open-source platform work, most prominently nSelf, a self-hosted backend CLI that replaces managed databases with a single command on a $4-per-month server.
In 2024 he shifted focus to AI infrastructure as Principal AI Engineer at Unity. The current build is nClaw, a multi-model agent platform with hybrid retrieval (PostgreSQL plus pgvector plus tsvector plus Reciprocal Rank Fusion plus cross-encoder reranking via BGE-M3), a Multi-AI Router that ladders across Claude, GPT-5.4, and local Qwen3 models on Apple Silicon, and persistent agent memory in Postgres that recalls year-old conversations.
Work Experience
12Principal Engineer at Unity since 2014. Last decade: enterprise contracting, distributed systems, open-source platform work (nSelf CLI). Last two years: Principal AI Engineer building nClaw, hybrid retrieval, and the Multi-AI Router pattern.
My Projects
12Open-source maintainer across nSelf CLI, ClawDE/nClaw AI engineering platform, Unity Sites, plus 13 published npm packages spanning solar position algorithms, Hijri calendar adapters, and lunar phase tools.
Blog
View all (17) →Infinite Memory: How nClaw Recalls a Conversation From a Year Ago
The context window is not memory. It's working memory. Real recall across months requires server-side thread storage, a memory extraction pipeline, summary pyramids, an alias matrix, and hybrid retrieval. Here's how nClaw builds all of it.
ReadOwning Your Stack: Local LLMs Plus Frontier Models, On Your Terms
A privacy-first architecture for running sensitive AI workloads locally on Apple Silicon while selectively routing to frontier models, with a redaction pipeline, cost breakdown, and honest assessment of where local inference still falls short.
ReadThe Multi-AI Router: Stop Sending Every Question to GPT-5.4 or Opus 4.7
Every request to a 70B frontier model that could've been handled by a 7B local model is money you didn't need to spend and latency you didn't need to add. The Multi-AI Router pattern fixes that by treating model selection as a first-class architectural concern.
ReadTestimonials
Aric delivered production-quality code on a tight deadline without cutting corners. He understood the system architecture quickly and shipped features that just worked. One of the most reliable engineers I have worked with.
James Whitfield
Engineering Manager at Apple
We brought Aric in as lead developer for a complex data visualization project. He took ownership from day one, made smart technology choices, and communicated progress clearly. The client was thrilled with the result.
Rachel Nguyen
VP of Product at Pearson
Working under Aric at Capgemini was a turning point in my career. He set high standards for code quality and took the time to explain the reasoning behind architectural decisions. A great mentor and a strong technical lead.
David Park
Senior Developer at Capgemini
Get in touch
Open to new projects, opportunities, or just a conversation.
