Led · Equity · Aug 2026

Oshen develops 1.2-metre C-Star unmanned surface vehicles that are wind- and solar-powered, can operate autonomously for months, and send data by satellite. The robots are light enough for a single person to launch from a boat and are designed to operate in coordinated groups that can cover hundreds of miles or concentrate at specific locations. Oshen’s C-Stars were the first uncrewed vehicles to collect real data from inside a Category 5 hurricane, and the company’s longest mission lasted eight months at sea. Manufacturing is now based in Plymouth and output has scaled from 15 units produced over the company’s first three years to a cadence of 15 robots every six weeks, with 100 additional units ordered. Customers and collaborators include the US Navy’s Naval Meteorology and Oceanography Command and partners such as ZeroUSV and MarineAI on defence-related trials. The company has grown from seven employees in September 2025 to almost 30 employees as it pursues higher-volume, lower-cost deployments to compete with larger rivals like Saildrone, Sofar Ocean and Boeing’s Liquid Robotics.

Led · Seed · Nov 2025

Emm is a UK-based femtech startup that has engineered what it calls the world’s first smart menstrual cup, embedding ultra-thin sensors into medical-grade silicone to collect real-time menstrual data. The companion app encrypts and stores this information, allowing users to track patterns and potentially flag conditions such as endometriosis, which currently takes an average of seven to ten years to diagnose. Founder Jenny Button conceived the idea during the COVID lockdown after noting the lack of reproductive health wearables despite widespread adoption of devices like the Oura Ring and Whoop band. Following thousands of prototypes and extensive user testing, the company is preparing for a full UK launch next year and has amassed more than 30,000 pre-orders. Looking ahead, Emm plans to expand into the U.S. market by early 2027 and sees opportunities to move beyond menstrual tracking into diagnostics, broader digital care tools, and even therapeutics. The startup recently secured $9 million in seed financing, providing runway for product launch, continued R&D, and market expansion.

Participated · Series A · Apr 2025

Neurolabs offers an Image Recognition as a Service platform that enables field teams to capture store-shelf photos on smartphones and receive near-instant, product-level insights powered by proprietary Visual AI and synthetic-data pipelines. The technology is positioned as a faster, lower-cost alternative to manual retail audits and integrates into enterprise retail tech stacks as a scalable infrastructure layer. Neurolabs says its platform has helped clients cut field operational costs by up to 32% and reduce product onboarding time by 93%. Clients cited in the article include a US soft drinks company, a major European beverage brand, and a UK-based manufacturer. The company is growing its computer-vision engineering team and deepening its commercial footprint across the UK, European, and US markets with plans to scale globally. Management frames the opportunity as addressing an $800 billion problem of inefficiencies in CPG supply chains. Neurolabs offers a no-code/low-code synthetic computer vision platform that helps retailers deploy automation solutions at substantially lower development and deployment cost. The company focuses on precise 3D modeling of everyday physical objects and is compiling a retail 3D repository of roughly 100,000 digital twins of consumer packaged goods. Those digital twins are intended to enable synthetic computer vision across the CPG lifecycle, from manufacturing and distribution to in-store, e-commerce and recycling. CEO and founder Paul Pop highlights the technical challenge of anticipating and reproducing real-life changes in packaging and design, which the company addresses with its synthetic computer vision approach. Neurolabs plans to scale operations and expand its offering to include several consumer packaged goods use cases. Financially, the company raised a €3 million seed round and has raised a total of $4.9 million since 2019. Neurolabs builds a platform that trains object-recognition computer-vision models using solely synthetic (computer-generated) data and delivers models via server deployment or downloadable packages accessible through API calls. The company targets industries such as food service, grocery retail, automotive, smart city, healthcare, manufacturing and agriculture, with many deployments to date in food service for autonomous checkout. Neurolabs claims synthetic-data approaches enable much faster and cheaper model development (up to 20x faster and 100x cheaper versus traditional methods) and offers a no-code interface for RPA/software developers as well as virtual data-generation tools for CV engineers. The startup was founded in August 2018 and has offices in Edinburgh and Cluj-Napoca; it currently has about 10 employees and planned modest hires by end of 2021. Financially, the company reported current revenues under €250k and said it was targeting €1M in revenue by Q1 2022; most of its funding has been used to fund R&D and engineering of the Neurolabs platform.

Participated · Equity · Oct 2024

Runware offers an end-to-end infrastructure layer that lets enterprises add generative-AI capabilities—spanning image, video, and audio creation—through one unified API. By combining its Sonic Inference Engine hardware with optimized software, the company targets up to 10× lower costs and markedly reduced latency compared with standard GPU-based providers. Current customers include Wix, Together.ai, ImagineArt, Quora, and Higgsfield, among others, highlighting traction across both consumer and enterprise workflows. The platform addresses three pain points in media AI workloads: fragmented model access, performance latency, and poor unit economics at scale. Runware’s vertically integrated approach aims to remove the need for clients to integrate dozens of vendors or negotiate large volume commitments, accelerating time-to-market for new AI features. With offices in London and San Francisco, the company plans to channel new capital into expanding its team, enhancing the Sonic Inference Engine, and broadening model coverage. Although no revenue figures were disclosed, management reports repeated usage growth peaks as customers roll out new AI features on the platform.

Participated · Series A · Sep 2024

DeepOpinion builds an agentic, no-code AI platform that automates complex, knowledge-intensive back-office workflows by leveraging context-understanding capabilities and large language models. The platform adapts to any document type, connects to over 200 enterprise software tools, and uses self-optimising AI agents that learn from human feedback to enable touchless end-to-end processing. In insurance scenarios the company says it can automate claims in about 90 seconds versus human processing times of roughly five weeks after disasters. DeepOpinion's technology targets unstructured data from emails, documents, messages and forms to scale automation where traditional tools fall short. Clients across insurance, financial services and telecommunications include Siemens, e&, BitPanda, HannoverRE, Uelzener and Allianz. The company raised €11 million in a Series A to fund global expansion and further development of its core AI platform.