About us

Anthus is an AI engineering company that builds and operates self-aligning AI systems — custom models, agent harnesses, and evaluation loops with a human in the loop — for teams that need them to keep working in production.

About AnthusIn 2009, Ryan Porter developed an innovative event ticket sales system for a Tiësto world tour, sparking a business that generated up to $64 million annually since 2007. We came together as a team as the business expanded, going from a tiny startup to a department within a large multinational corporation, Tao Group Hospitality. We have worked cohesively for more than a decade, safeguarding an impeccable record in reliability and security.

Throughout our journey, we not only integrated various third-party platforms—including Salesforce, Marketo, Salesforce Marketing Cloud, Mailchimp, and Paytronix into our system but also successfully navigated through multiple technological revolutions. From embracing cloud computing to harnessing serverless architectures and deploying artificial intelligence applications, we continuously adapted and innovated while maintaining operational excellence and security. Even amidst rigorous audits from top-tier firms and scrutiny from our parent company, Madison Square Garden Entertainment, our innovative approaches to service management maturity and risk management consistently earned approval.

Our endurance since 2007 is a testament to our commitment and capability in preventing business interruptions arising from downtime, software malfunctions, security incidents, or data losses. The hackers never got us, despite being a prime target. When a catastrophic failure struck an entire AWS data center, we relocated the entire system to a new data center within hours, ensuring that ticket scanning at nightclubs proceeded without a hitch. We have navigated through massive DDoS attacks, primary database server failures, silent failures in third-party systems, and everything else you could imagine. We always kept the revenue flowing. Then we smoothly handed it off to the next team with no business interruption or impairment.

Our Mission

Deliver reliable, secure, and efficient business solutions using collaboration between human and artificial intelligence in every aspect of development and operations. We build systems that keep working when nobody is watching.

Our Values

  • Prioritize Solutions Over Tools

    Investing in products and services only delivers business value if you're in the business of products and services. We're in the business of solutions.

  • Focus on Business Logic

    The only code you should be writing is the business logic that solves real problems. Don't waste time reinventing wheels.

  • Continuously Improve

    Enable rapid, iterative change through CI/CD and DevOps—then let the systems improve themselves. Self-optimizing classifiers and self-steering agent systems get better from production feedback without waiting for an engineer to notice.

  • Implement Infrastructure as Code

    Leverage DevOps to implement Infrastructure as Code, and MLOps and LLMOps to do the same for the models. Every part of a production system—including training runs and evaluations—should be created by code so it's reproducible, not clicked together by hand.

  • Commodify AI Models

    Treat AI models as replaceable, not magic black boxes. In a world with no moats, don't invest too much in any given castle.

  • Optimize Resource Usage

    Balance efficiency with cost-effectiveness. When intelligence is cheap, the goal shifts from conserving compute to conserving context and cognitive load.

The night we stopped being the loop

When we started using AI to write the code and make the decisions, the chat way of working broke under the volume. A person had to sit there keeping the agents going or the whole process stopped, and a person had to hold in their head what was happening across every parallel session in every project for every client, because nothing else did. The low point is on the record, like everything else we do.

So we stopped being the loop. Tactus names the three ways a person can be in the loop, supervised, unattended, and asynchronous, and makes the third one something you can program, so an agent runs on its own and interrupts a human only at a decision point. Kanbus is the memory that lets work move between Claude Code, Codex, or whatever comes next, because we've held since 2023 that models are commodities. Plexus is the scorecard where reviewers correct a decision model's verdicts and their explanations become policy. We ran all of it on our own work first, in public.

How We Build Today

The AI era doesn’t remove the need for operational excellence—it raises the stakes. We use a cybernetic approach: clear intent, layered safeguards, and production feedback loops that continuously harden the system. Delegating the work turned out to be easy; observing it is the whole job. Read more in Cybernetic Development.

Spinoffs

Running AI agents all day created a problem of its own: the bottleneck moved from writing the code to monitoring and steering the agents running it. Sitting in front of a terminal all day is not the answer. That workflow problem became a hardware venture. We spun up Anthus Microelectronics to build the desk displays that grew out of it — handbuilt microelectronics running a custom handbuilt (AI-assisted) OS, real and handmade even where AI accelerated the build.

The first product is B0rd: a standalone LED-matrix desk display for glanceable information — launch countdowns, agent status, notifications, an idle clock — that you can read from across the room instead of a screen you have to sit in front of. Matching units even stay in sync without pairing or a hub. See the B0rd solution page, visit b0rd.info for the product site, or browse the Etsy shop.

Who We Work With

  • Product and engineering teams putting LLM agents or classifiers into production, not just prototypes
  • Operations and QA organizations that need AI to make judgments at scale while experts stay in control (the Call Criteria pattern)
  • Businesses with revenue-critical systems where downtime, security incidents, and silent failures are unacceptable

Key Facts

Company name
Anthus AI Solutions
Type
AI engineering and operations company
Founder
Ryan Porter
Website
https://anth.us
Core offering
Self-aligning AI systems: custom models, agent harnesses, and evaluation loops with a human in the loop
Services
LLM fine-tuning, agentic AI, MLOps and LLMOps, serverless architecture, AI-assisted operations
Track record
Team together for over a decade; production platforms operated for 14+ years; Call Criteria in production with RLHF for two years
Platform
Plexus, an MLOps platform
Spinoff
Anthus Microelectronics (B0rd LED-matrix desk display)
Contact
https://anth.us/ryan

Frequently Asked Questions

What does Anthus build?

Production AI systems that improve from feedback: fine-tuned classifiers, agent systems, and the evaluation and MLOps infrastructure around them. We also operate what we build.

How is Anthus different from a typical AI consultancy?

We are operators first. The team has run a revenue-critical platform since 2007, through data center failures and DDoS attacks, before applying the same discipline to AI. We ship systems with human-in-the-loop feedback, not demos.

What is human-in-the-loop AI?

Experts review and correct the AI's judgments, and those corrections become training signal. In Call Criteria, QA specialists guide the models this way, and the system has improved through reinforcement learning from human feedback (RLHF) for two years.

Do you only work with one model provider?

No. We treat models as replaceable components and design systems so a model can be swapped without rebuilding the business logic.

Now, we bring our depth of experience and technical agility to your projects. What can we develop and operate for you?