At 007 Venture Partners, we believe the highest-impact investments tackle large, structural inefficiencies rooted in real-world constraints. One such challenge often ignored even within AI discourse is the persistence of manual visual inspection across global manufacturing.
Despite decades of automation elsewhere, today ~95% of manufacturing vision inspection tasks remain human-run slow, costly, and inconsistent. Allus AI is fixing that by building the universal vision intelligence layer that finally brings industrial-grade AI to every production line.
The Trillion-Dollar Problem: Manual Vision Inspection at Scale
Global manufacturing is a massive economic backbone, but its quality control workflows are stuck in the past. Traditional machine vision systems require months of custom engineering, are expensive to build and maintain, and often fail outside narrowly defined use cases.
The result? Most factories still rely on human inspectors to catch defects, leading to:
- Escalating labour costs
- Quality inconsistency across shifts and sites
- Bottlenecks in scaling production
This is a structural bottleneck that affects profitability, competitiveness, and operational safety in factories worldwide.
Allus AI Is Built to Change That
Allus AI replaces slow, bespoke vision systems with a foundation-model driven platform that can be configured in minutes, not months.
Instead of handcrafted rules for each inspection task, Allus uses a 1B-parameter vision foundation model trained on over 1.5 billion proprietary industrial data pairs—data collected over five years across live production lines. This allows the system to interpret factory scenes the way humans do but at machine speed and precision.
By enabling inspection workflows via natural language prompts and a handful of reference images, Allus slashes deployment time and cost while unlocking automation for both global enterprises and mid-market manufacturers.
From Research Labs to Real Factory Floors: The Founders Behind the Vision
The Allus founding team: Kai Cui (CEO), Shijie Wang (CTO), and Jason An (COO) brings together rare technical excellence and lived industrial experience.
- Kai Cui grew up embedded in the manufacturing ecosystem through Foxconn, where he managed on-ground production operations. This gave him deep domain intuition about real pain points that industrial AI must solve.
- Shijie Wang is a former NVIDIA LLM scientist, driving the development of Allus’s vision foundation model architecture and training processes.
- Jason An has extensive experience deploying AI in production-grade industrial environments, ensuring Allus technology works not just in theory but at scale, on real factory floors.
Together, they’ve built what no one else has: a truly generalised, adaptable vision abstraction layer for manufacturing, analogous to what large language models did for text.
What Makes Allus AI Different
Allus AI replaces slow, expensive, and highly manual vision inspection systems with a rapid, AI-native platform that can be deployed in minutes rather than months. At a time when ~95% of manufacturing vision tasks remain manual, Allus delivers automated quality inspection at 93% lower cost and ~26,000x faster deployment than traditional solutions, unlocking automation for both large enterprises and mid-market factories that were previously priced out.
Traditional industrial vision systems are:
- Custom-engineered per task
- Slow to deploy
- Brittle outside narrow use cases
Allus AI’s approach is different:
- 10-minute deployments vs 3–6 months of traditional integration
- No rule-based modelling, just natural language + reference images
Compounding Data Advantage
While general foundation model players like OpenAI and Google lack access to real factory data, Allus has built proprietary industrial data at scale. A moat that is inaccessible to competitors because manufacturing environments are secretive, locked down, and difficult to instrument.
Every deployment feeds more real-world data back into the model, strengthening performance and customer lock-in.
Strong Traction, Early Product-Market Fit
Allus’s traction is exceptional for an industrial AI company. Within months of launch, it secured contracts from global leaders including Apple, Tencent, Panasonic, Tesla, KUKA Robotics, Corning, Novo Nordisk, and Mondelēz. This is a level of early commercial adoption that is rare even among later-stage industrial AI companies.
A Convergence Moment
Manufacturing stands at an inflection point. Labour shortages, wage inflation, aging inspectors, supply-chain pressures, and rising quality standards are driving factories to rethink inspection systems. Meanwhile, advances in foundation vision models, edge and cloud compute, and real-world data capture finally make universal AI vision viable.
Allus sits at the intersection of all these trends.
Advancing Industrial Competitiveness and Inclusion
At 007 Venture Partners, we invest not just for returns but for impact and transformation.
Allus AI:
- Reduces waste and defect rates
- Improves yield and throughput
- Addresses skilled labour shortages
- Makes advanced automation accessible to mid-market factories previously priced out
In doing so, it accelerates productivity and competitiveness across global manufacturing.
Why We Invested
Allus AI exemplifies the type of company we seek at 007VP:
- Founders with rare market and domain fit
- Massive, structural market opportunity
- Defensible, data-driven moat
- Tier-1 validation and strong early traction
This was an oversubscribed seed round, joined by Tencent, Pioneer, Transpose, Mixpanel CEO Suhail, ex-Shopify COO / Opendoor CEO Kaz, and Y Combinator.
We’re proud to back Allus as it scales towards becoming the universal vision intelligence layer for global manufacturing.
Looking Ahead
With a clear runway, a compounding data advantage, and strong strategic interest from industrial automation leaders, we believe Allus has the potential to redefine how factories see, inspect, and automate quality at scale.
We’re excited to support this team as they lead the next wave of industrial AI innovation.
