Where research meets real operations
These are the areas where our experience runs deepest, from the science behind a technology to the operations that have to run it. They describe professional expertise, not a catalogue of packaged services. Each example is labelled to show whether it comes from the founder’s career before Parseus or from work at Parseus AI.
Enterprise AI and agentic AI
Putting models and AI agents to work inside real business processes, with clear boundaries, oversight and accountability, so that automation scales without the judgment leaving people’s hands.
What it involves
- Deciding which decisions an AI system may make, which it may recommend, and which stay with people
- Designing multi-agent workflows where each agent has a defined responsibility
- Making reasoning visible and keeping a record of what each system did and why
Experience behind it
- Founder’s career Image-safety systems at Amazon, where the machine handled scale and a person made the consequential calls
- Parseus AI Maestro AI, a released multi-agent application with separate agents for finding, tailoring, critiquing and fact-checking
- Parseus AI Agent governance tools currently in development (see Projects)
Physical AI and robotics
Perception and decision systems for machines that act in the physical world, where light, heat, dust and motion never match the lab, and where a mistake has physical consequences.
What it involves
- Perception that stays reliable under harsh, changing conditions
- Finding and preparing for the rare situations that decide whether a system is safe
- Keeping operators informed and in control of machines that act on their behalf
Experience behind it
- Founder’s career A vision system that guides welding through the glare of the arc (Novarc)
- Founder’s career Perception work for driving systems, focused on rare and unseen situations (Matt3r, ongoing)
Manufacturing and industrial automation
Connecting data, equipment and people on and around the factory floor so that operations can anticipate problems, plan on evidence and run with visibility at every level.
What it involves
- Bringing scattered production data and knowledge into one dependable picture
- Planning and scheduling grounded in evidence rather than guesswork
- Automation that fits the realities of existing lines and the people who run them
Experience behind it
- Founder’s career Connecting and centralizing data and knowledge across circuit-board manufacturing sites (Summit)
- Founder’s career Vision-guided welding on the production line (Novarc)
Intelligent systems and computer vision
Vision and sensing systems that hold up outside the lab: through dust, glare, vibration and weather, and at the scale of millions of images.
What it involves
- Designing for the conditions the system will actually face, not the demo
- Measuring performance against real outcomes, including the cost of a miss
- Deciding where human review belongs in the loop
Experience behind it
- Founder’s career Computer vision for mining, where a missed detection can jam a crusher (Weir Motion Metrics)
- Founder’s career Welding vision through arc light (Novarc)
- Founder’s career Unsafe-image detection at Amazon scale
Enterprise architecture and operating models
The structures, data flows, ownership and ways of working that let AI become part of how an organization actually runs, rather than a project on the side.
What it involves
- System and data architecture that connects AI to the operations it supports
- Clear ownership, accountability and escalation for automated decisions
- Operating models that keep engineers, operators and decision-makers working from the same picture
Experience behind it
- Founder’s career Operational systems across hundreds of sites at Amazon
- Founder’s career Building a shared data and knowledge backbone for a multi-site manufacturer (Summit)
Strategy, implementation and scaling
Choosing where AI is worth it, proving it in a pilot that reflects real conditions, and taking it to production without losing reliability along the way.
What it involves
- Separating what new research enables from what a headline claims
- Defining pilots with measurable success criteria and a path to production
- Scaling from “it works once” to “it works every time”
Experience behind it
- Founder’s career Turning research into scale at Intel
- Founder’s career Fractional technology executive roles with startups (Novarc, Matt3r)
- Founder’s career A path from academic research, through a startup, to large corporations
Start with a conversation
We offer ten complimentary 20-minute AI strategy conversations by application, and monthly AI roundtables on making AI work in practice. For anything else, email us.