Case studies

One thesis, shipped seven ways.

Notum builds the discipline and autonomy layer that turns raw intelligence (an LLM, a GPU fleet, a robot, a sensor) into a system you can actually deploy. DogOS does it for robots. DoggieHarness does it for coding agents. ClusterFlock does it for compute. nMesh does it for communications. The invoicing and HR platform and the driver-assist system do it for business and consumer workflows. The bet is consistent: intelligence is becoming cheap; the reliability layer around it is the durable value.

ClusterFlock

Every GPU you own, one AI backend.

A distributed AI compute orchestrator that unifies DGX, RTX, Apple Silicon, and Jetson hardware into one self-healing cluster behind an OpenAI-compatible API, with autonomous multi-agent missions on your own metal. Open source under the MIT license.
- AI infrastructure
- Open source · MIT
- v0.7.10
Read the case study
DoggieHarness

Weaker models, coding like stronger ones.

A coding harness that wraps any LLM in a discipline layer it does not have on its own: self-audited file edits and independent work validation. It lifts the floor, so smaller and cheaper models produce edits that would normally require a frontier model.
- AI tooling
- Coding agents
Read the case study
nMesh

Networks that survive the disaster.

A self-organizing, low-energy mesh network built on LoRa radio: no towers, no fixed infrastructure, battery life measured in months to years. One platform for disaster-zone communications and city-scale sensing.
- LoRa mesh
- Emergency response
- Smart cities
Read the case study
Invoicing & HR AI platform

Back office that runs itself.

An AI platform that automates the judgment in back-office work, not just the data entry: invoices read, matched, and routed with humans touching only exceptions; HR onboarding, records, and routine questions handled automatically.
- SaaS
- Finance & HR automation
Read the case study
In-car driver-assist AI

Driver intelligence for the cars people actually drive.

A self-contained driver-assist AI that needs neither CarPlay nor Android Auto: real-time road and hazard awareness, contextual alerts, and driver-state monitoring for the existing vehicle fleet, not just new dashboards.
- Edge AI
- Automotive
Read the case study
Nowcast storm predictor

The ninety minutes big forecasts miss.

A custom nowcasting model for short-range, localized storm prediction: what the sky above this site will do in the next hour, precisely enough to act on. Active research, with a direct roadmap into weather-aware autonomy.
- Ongoing R&D
- ML forecasting
Read the case study
Autonomous site monitoring

Autonomous inspections on a live construction site.

A Massachusetts construction company needed continuous, comparable site intelligence without putting more people in harm's way. DogOS now runs their inspections autonomously: identical shots from identical positions, hazards flagged between walks, no operator on site.
- Physical AI
- Construction
- Live since 2024
- DogOS
Read the case study

Building something in this shape?

Tell us about it. If it's complex, we're interested.

Start a project

ClusterFlock

DoggieHarness

nMesh

Invoicing & HR AI platform

In-car driver-assist AI

Nowcast storm predictor

Autonomous site monitoring

Building something in this shape?