The most expensive part of training a computer vision model is not the compute. It is the human being who has to sit there and draw boxes around thousands of cans of soup, labeling each one. Neurolabs, a London and Edinburgh startup, decided to skip that person entirely. For the last six years, the company has been building a library of over 100,000 three-dimensional digital twins of consumer packaged goods, from cereal boxes to shampoo bottles, and using them to generate perfectly annotated, synthetic shelf images [SignalBase]. This is the engine for what they call Execution Intelligence, a platform that promises to turn a field rep’s smartphone photo into real-time, SKU-level insights on pricing, promotions, and out-of-stocks [neurolabs.ai].
The synthetic data wedge
Neurolabs’s core bet is that synthetic data is not just a cheaper way to train AI, but a better one for the specific chaos of retail. Traditional shelf auditing relies on either manual human checks,slow, expensive, inconsistent,or computer vision models trained on photos of real stores. The latter requires collecting and labeling a massive, diverse dataset for every new product, a process that can take weeks and still struggle with odd angles, poor lighting, or new packaging. Neurolabs flips the script. Its team creates a 3D model of a product using a few reference images, then uses that digital twin to generate millions of training images in countless virtual store environments [startupintros.com]. The result, the company claims, is a system that can recognize a new SKU in days, not months, and with “superhuman” accuracy because the synthetic data has perfect, pixel-level annotations no human could match [Gust].
Their product, ZIA (Zia Intelligence Automation), is sold as an Image Recognition as a Service platform. It gives CPG brand managers and field teams an API and a dashboard to upload photos and get back analytics: share of shelf, planogram compliance, promotional execution, and competitive benchmarking [eutechfuture.com]. The commercial argument is straightforward. It turns a subjective, episodic audit into a continuous, quantified data stream, all using hardware the team already carries in their pocket.
From generalist to focused execution
Neurolabs did not start life squarely aimed at the grocery aisle. Founded in 2018 by computer scientists and mathematicians Patric Fulop, Paul Pop, and Remus Pop, the company initially offered a broader computer vision platform where developers could build custom object-recognition models, with early use cases in foodservice and manufacturing quality control [BRAND MINDS]. The pivot to a dedicated retail and CPG execution layer appears to be a classic case of finding the wedge. The pain point is acute, the budgets exist, and the synthetic data approach directly attacks the biggest bottleneck to scaling image recognition across millions of SKUs in thousands of stores.
The founding team, who studied together at the University of Edinburgh, now fill the classic C-suite roles: Paul Pop as CEO, Patric Fulop as CTO, and Remus Pop as Chief Revenue Officer [TFN]. Their backgrounds in finance and engineering at firms like J.P. Morgan and Bank of America Merrill Lynch suggest a commercial and operational discipline that aligns with selling into large, slow-moving enterprise clients [TFN].
Traction and the competitive shelf
Neurolabs has raised a total of $12.45 million to date, with a $7.8 million Series A round led by Nauta Capital that closed in April 2025 [Caplight, April 2025] [nautacapital.com, 2026]. The company has scaled to between 101 and 250 employees, indicating a significant investment in both R&D and go-to-market [Crunchbase, 2026]. A key recent move was the launch of its Execution Intelligence dataset on the Databricks Marketplace in May 2026, a partnership that positions its analytics as a native component in the modern enterprise data stack [streetinsider.com, 2026].
They are not alone on the shelf. The competitive set includes well-funded players like Trax and Vispera, which also use computer vision for retail, and Pensa Systems, which employs AI for inventory intelligence. The table below outlines the competitive landscape Neurolabs operates within.
| Company | Primary Approach | Key Differentiator |
|---|---|---|
| Neurolabs | Synthetic data & digital twins | Rapid SKU onboarding, claims superhuman accuracy from synthetic training |
| Trax | Computer vision (real-world imagery) | Extensive market footprint, retail-focused SaaS platform |
| Vispera | Computer vision (real-world imagery) | Strong focus on shelf execution and compliance for CPG brands |
| Pensa Systems | AI & computer vision | Emphasis on inventory intelligence and out-of-stock prediction |
Where the model could misfire
The risks for Neurolabs are not technological fantasies but practical go-to-market pressures. The company’s entire value proposition hinges on a few key assertions that remain largely self-reported.
- Accuracy in the wild. Claiming “superhuman” vision is a bold marketing stance [Gust]. The real test is performance on messy, real-world shelf images from a dusty corner store under fluorescent lights, not in a controlled virtual environment. Any significant error rate in identifying SKUs or counting facings would immediately undermine trust with brand managers whose bonuses depend on this data.
- The cost of building twins. While adding one SKU may be fast, building the initial library of over 100,000 digital twins represents a colossal upfront investment in artist and engineer time [SignalBase]. The question is whether this library constitutes a durable moat or simply a high barrier to entry that well-funded incumbents could eventually replicate.
- Sales cycle gravity. Selling enterprise SaaS to global CPG companies is a marathon. The sales cycles are long, procurement is byzantine, and incumbents have deep relationships. Neurolabs’s team has finance experience, but scaling a direct sales force to compete with Trax’s established footprint is a capital-intensive grind.
The company’s answer to these risks is its partnership play, like the Databricks integration, which bypasses some traditional sales friction by embedding directly into the customer’s existing data workflow [LinkedIn, 2026].
The next twelve months
For Neurolabs, the immediate future is about proving that its synthetic data engine can drive not just pilots, but enterprise-wide deployments. Key milestones to watch will be the announcement of flagship CPG customers beyond early adopters, and any data on renewal rates and expansion within those accounts. Another likely move is a subsequent funding round to fuel the sales expansion needed to truly challenge the incumbents. The Databricks marketplace launch will be a critical leading indicator; significant uptake there would validate the product-led growth angle.
On the back of an envelope, the unit economics of this bet are fascinating. If a single field auditor costs a CPG brand roughly $50,000 per year in salary, travel, and expenses, and can visit a store once a month, the value of continuous, automated insight from every store visit is clear. For a brand spending $5 million annually on a thousand auditors, even a 20% efficiency gain represents a million dollars in saved cost or redeployed revenue opportunity. Neurolabs does not need to replace every auditor to win. It just needs to make each one ten times more effective, and charge a fraction of the saved cost for the privilege.
To succeed, Neurolabs must ultimately beat Trax at its own game: becoming the default, trusted source of truth for what’s actually on the shelf. It is betting that its synthetic data foundation makes it not just cheaper, but smarter and faster,a next-generation map for a territory that has always been frustratingly opaque.
Sources
- [SignalBase] Neurolabs company profile | https://signalbase.com
- [neurolabs.ai] Execution Intelligence for CPG Commercial Teams | https://www.neurolabs.ai
- [startupintros.com] Neurolabs company profile | https://startupintros.com/orgs/neurolabs
- [Gust] Neurolabs company profile | https://gust.com/companies/neurolabs
- [eutechfuture.com] Neurolabs: Pioneering Synthetic Data and Visual AI | https://eutechfuture.com/deep-tech-in-europe/neurolabs-pioneering-synthetic-data-and-visual-ai-to-transform-retail-execution/
- [BRAND MINDS] Innovative computer vision startup Neurolabs has attracted $1.2M in funds | https://brandminds.live/innovative-computer-vision-startup-neurolabs-has-attracted-1-2m-in-funds/
- [TFN] Neurolabs team background | https://tfn.com
- [Caplight, April 2025] Neurolabs funding information | https://www.caplight.com/company/neurolabs
- [nautacapital.com, 2026] Nauta Capital leads Neurolabs Series A | https://nautacapital.com
- [Crunchbase, 2026] Neurolabs headcount data | https://www.crunchbase.com/organization/neurolabs
- [streetinsider.com, 2026] Neurolabs launches on Databricks Marketplace | https://www.streetinsider.com
- [LinkedIn, 2026] Neurolabs Ontology and Databricks partnership | https://www.linkedin.com/company/neurolabs