Ain’t a conundrum: why we invested in

At Speedinvest i — the newest focus fund of Speedinvest dedicated exclusively to Industrial Tech — we believe that production sector is leading the way in the application of artificial intelligence technology. To dramatize even more, studies like “The State of European Tech Report” (2018, Atomico) state clearly that traditional industries (such as industrial manufacturing) have the biggest influence on the European economy today (see the Figure 1 below). And since European growth has been rather flattening in the last quarters, we at Speedinvest i feel a great deal of responsibility to contribute to the European growth by investing in the best European industrial startups. ☺

Figure 1. “Tech” is the engine for the GDP growth: Share of total gross value added in Europe.

On that note, our latest investment in Conundrum reflects our ambition and convincement in the above presented hypothesis. This UK company focuses on research in the industrial AI and development of its proprietary (AI based on deep learning with automatic data preprocessing) technology and software products. Further, key technological features are: 1) both supervised and unsupervised learning for different types of tasks (multiclass classification, regression, anomaly detection and more), 2) adaptability (auto-retraining in case of equipment/machinery changes/modifications) and 3) scalability (ability to draw data from any number/type of sensors as well as industrial systems). For now, their advanced AI algorithms and platform are being developed for the two use cases: 1) Predictive Maintenance (PdM) and 2) Quality Control (QC). PdM and QC have become focus areas for many producing companies, due to promising potential increases in profit margins. However, benefits from being able to predict failures and automate the quality control go beyond obvious monetary savings.

Conundrum: Solution, Focus and Use Cases in Detail.

Conundrum’s current focus lies on industries such as pulp & paper, metal & mining, oil & gas and pipe welding. An ongoing maintenance of machinery and equipment greatly impacts a bottom line of operations in these industries. Studies prove that 42% of the unplanned downtime is caused by production asset failures (CIO Magazine). The prospects of reducing this number are immense. Only taking PdM into account, the potential savings in heavy industries are an increase of asset availability by 5–15% and a reduction of maintenance costs by 18–25% (McKinsey). This easily justifies Conundrum’s focus on PdM.

A number of industrial players have some sort of initiative or solution in place to achieve these potentials. According to Roland Berger (The study on Predictive Maintenance), around 81% of firms report that they are already now devoting time and resources to the topic of predictive maintenance. Looking into the future, PdM worldwide market will expand to around USD 11 billion by 2022 (see the Figure 2 below). This is a great looking playground for the startups like Conundrum!

Figure 2. Predictive maintenance: Forecast global market development to 2022 (USD billion).

However, PdM is only the first use case Conundrum addresses with its platform.Quality Controlis the second largest (in terms of estimated potential) use case, which gains in importance, due to increasingly shorter time-to-market deadlines and more demanding quality regulations. Such ML algorithms, as Conundrum’s, contribute to the growing automation of quality control in the industrial space, what enables faster feedback loops due to real-time analysis of gathered data, rather than reliance on humans for detection of errors.

But what kind of data do such digital platforms leverage, integrate and gain insights from? In case of Conundrum the data is drawn from various systems: 1) shop floor systems, such as MES (manufacturing execution systems), SCADA (supervisory control and data acquisition) or DCSs (distributed control systems); 2) enterprise systems (such as ERP, SCM etc.); as well as 3) industrial PCs and machines or sensor data from IIoT devices. Combination and aggregation of different sources leads Conundrum to get the most out of available data.


However, even technologically best solution does not presuppose a startup success. We like to say, “We invest first of all in the team”, which reflects our strong believe that a solid team is critical. In Konstantin, Vlad and Victor we found just that, a genuinely interested, passionate and energetic, group of data science experts and friends among each other, that brings (alongside with a team of 12 DL and DE specialists) unique domain expertise in industrial deep learning.

Technological Partner.

Conundrum has proved machine and deep learning domain expertise by sharing its knowledge, experience and cutting-edge ideas with leading universities and at a number of conferences in cooperation with NVIDIA Deep Learning Institute.Conundrum is a participant of NVIDIA Inception program as one of the AI-first world leading companies. Conundrum was selected as a top ten AI companiesand presented its platform at the GTC conference in Munich. Conundrum will likewise be a one of the speakers at GTC in San Francisco this year.

Forming an opinion.

On a side note, at Speedinvest i, we have screened a number of similar startups in the space, which aim to become the top go-to-address for industrial clients. Meanwhile, we have started forming an opinion about what are the necessary success criteria on the digital platform market. Wait up for our next blog post, where we will elaborate on that point in more detail!