A superior market research and monitoring tool that leverages the exponentially growing, publicly available information, is sought after by investors and companies alike. With STARTUP RADAR we want to automate the first steps of every tech due diligence and commoditize it. We believe this will inevitably benefit Berlin, London, Barcelona, etc. (where investors may have their personal network), or traditionally overlooked markets, like women-led startups. It will also provide a levelling field for investors, end-user and clients (particularly angels and early-stage VC, who cannot pay for the kind of manual-intensive due diligence that Private Equity can afford.
Technically, a first step in facilitating the automatic processing of natural language (NLP) is to extract structured data from text. Hence one of the key steps is extracting semantic relations between entities: Relation extraction (RE). For our business exploitation use case, we aim at detecting:
• Mergers & acquisitions
• Funding of startups
• Launch of products
1. To automate the first steps of every tech due diligence and commoditize it.
2. To provide a levelling field for investors, our end-user and clients.
3. To facilitate reporting and benchmarking through dramatically improving the “Similar To” feature, and its “Explainability” of suggestions/recommendations.
We propose scalable models based on neural retrieval. We will also use it for explaining the user's recommendation (Explainable AI).
We will exploit both the structured & unstructured info about the companies as item attributes. Using our biz2vec embeddings, we will identify which docs share more coincidences and produce meaningful explanations. By combining both discrete and semantic relatedness justifications on our outputs, the user will assimilate our results more naturally.
In terms of market impact, the growth of the venture capital investment market in the world over the last decade is shocking, reaching $621B in 2021 from $12B in 2010. Nearly x50, with 111% YoY in 2021. And yet, there is very limited amount of software used to streamline these transactions. It still is network-dependent (who knows whom) and inefficient (deals take months to be scouted, vetted and closed).
The KERs obtained during the experiment will provide a revenue path for the lead SME (LKN). The exploitation plan has its foundations in the alignment with LKN technical expertise and core business. LKN current product portfolio shares strategy and distribution with SR, whereas the end-users and clients (i.e. investors) mean an expansion in LKN market.
KPI01: STARTUP RADAR (SR) makes use of a dataset from EUH4D (pending authorisation), from the JRC.
KPI02 and 05: One dataset to be open sourced, and a clear revenue stream identified.
KPI08: KERs are the outcome of 4 use-cases with clear owners, emphasizing the short-term end user client, i.e. early VC investors. The work methodology puts KPI08 at the center.
KPI15: Aim at generating 3 job positions in 2023 and 6 in 2024.