We build AI systems that make accurate visual judgments in the field, on the phones and edge devices your crews already carry. Rail, utilities, construction, manufacturing. Offline. Deployed in weeks. PhD-led, Tokyo and Osaka.
Field work asks your crews to make visual judgments all day long. Is this weld sound? Is that clearance within spec? Was this connection made correctly before we energised the line? Today, most of those judgments rely on a senior inspector's eye, a tape measure, and a paper checklist. We turn them into AI that runs on a phone.
Catch the defects, wear, and damage that matter - including the rare ones a traditional dataset would never cover. Track condition over time. Prioritise what needs a human second look.
Dimensions, clearances, alignment, wear depth. From a smartphone camera, without a total station, without shutting down the asset.
Pass/fail checks before a crew energises, lifts, cuts, or closes up. The inspector's judgment, captured once and available to every worker on site.
Vision-language models learn the task from a handful of examples, synthetic data covers the long tail, and the result distils down to a model that runs fully offline on hardware you already own.
Built for:
clearance and dimensional checks, wear detection, overhead equipment condition
live-line connection verification, pole and tower condition, vegetation risk
progress measurement, safety compliance, defect logging
surface defects, assembly verification, the long tail of "we'd check it if we had time"
The defects that matter most are the ones you've barely seen. Ten examples, not ten thousand. Traditional computer vision stalls here - we don't. Vision-language models learn the task from a handful of seed examples, and we generate synthetic data to cover the long tail of rare cases. Your dataset problem becomes a scoping conversation, not a two-year data collection project.
The pilot worked on a workstation. Then it needed a GPU server, a cloud account, a network exception, and six months of IT review - and the crews who were supposed to use it never got it in their hands. Our models run offline on the phones and edge hardware your teams already carry. Deployment is a download, not a procurement cycle.
You know the work matters. You also know no one is going to approve a large AI project on a hunch. You need evidence and a plan, but the larger players aren't interested.
We run a free two-week scoping engagement - normally valued at ¥800,000 - where we work with your team to define the problem, assess feasibility, and map out what a pilot would actually look like. No commitment beyond the two weeks. At the end, you'll know whether this is worth taking further, and we'll both know whether we're the right team to do it.
¥800,000
Free
Former strategy consultant at Strategy& (PwC) and Westpac Banking Group. Scaled a SEA start-up insurer from zero to 150M in insured exposure in two years, leveraging automation and AI.
15 years building autonomous systems across Amazon (Astro), Google Wing, and Figure AI. Full-stack autonomy from sensor fusion and perception through to real-time decision making.
Former consultant at KPMG with operating experience across APAC and the US. Leads partnerships and community, working with partners and enterprises to deliver AI engineering projects.
Applied research and real-world deployments. See how we bridge the gap between what AI promises and what it delivers in the field.
We're hiring with 5 open positions available. Come work on problems that matter, at the intersection of AI capability and APAC business reality.
Drive partnerships and client relationships across rail, utilities, construction, and manufacturing. Help teams understand what field visual AI can do for their operations and turn scoping conversations into deployed solutions.
Build and ship vision models that run on phones and edge devices. Work across the full pipeline: synthetic data generation, vision-language model fine-tuning, model distillation, and on-device deployment.
Don't see your role listed? We're always looking for exceptional people. Tell us what you'd bring and how you'd like to contribute.
