Data science startups review #15

September 14th - September 27th: computer vision in the offline world 💅🏗️ DS stack 💻; open source, pivoting from B2C to B2B, and other themes

Highlights

• Demo days and new cohorts from the leading accelerators; 

• From creating 3D models of nails (MainMe) to digitising construction sites (AirWorks) - use-cases for computer vision; 

• Open source in data science (Jina AI); Data science startup pivot - from B2C to B2B (Bayes Holding); 

• A hypothesis - those who have more data acquire those who have less (Snyk and DeepCode).

***
The last couple of weeks brought us demo days/announcements from three leading accelerators, Alchemist, Berkeley SkyDeck, and Techstars' Comcast NBC Universal LIFT Labs. Some companies from these accelerators are highlighted below, also please have a look here, here, and here respectively for the full lists of startups.


Data science stack/related 🥼💻

• Rectify, '... a tool meant to automatically detect and redact sensitive information when sharing documents outside of an organisation'; 

• Veamly, '... a "unified inbox" feed for developers that brings threads and messages from Slack, GitHub and Jira into one view...'; 

• Radicalbit, a DataOps platform, raised an undislosed round. 
  

Digital world/metaverse 🌎

• Sentinel, which is developing '... a detection platform for identifying synthesized media (aka deepfakes), has closed a $1.35 million seed round'📰; 

• Jina AI which builds neural networks powered search platform raised $5.5M. It's an open source startup with an interesting founder behind it, Han Xiao is the creator of bert-as-service and Fashion-MNIST. You could read about why he wants to work on the search challenge here 🔎; 

• Bayes Holding, a esports data company, has secured $6 million round. Behind this is a story of a pivot from a B2C gaming tool to a B2B esports data company 🎮.

World of atoms ⚛

• Percepta builds '... a computer vision technology to analyse existing security camera feeds and alert employees of shoplifting incidents'👮‍♂️; 

• AirWorks uses computer vision to create '... a way to convert aerial data into CAD drawings', raised $2.7M seed round🏗; 

• ManiMe 3D scans your nails to custom fit gel manicure💅. 

Exits 💸

• Apple acquiredScoutFM, a startup that uses AI '... to create automatically-curated lists of podcasts strung back to back to back to create a personalised talk track for your day' 🎧; 

• Snyk, a cybersecurity startup, acquiredDeepCode, the ' Grammarly for coders'. That's an interesting deal that might highlight how difficult is to build ML for coding 💻.

  Several startups who've attempted to do this, did not go far enough, even though they've approached the challenge from various angels. Prodo.ai seems to be closed, as well as Sourced, that previously raised $6M. Acquisition of

  DeepCode may shed some light on why it did not work out well for startups. One of the challenges for ML for coding startups was the limited access to high quality code to train models on. Even though open-sourced repos were available on GitHub, getting to high quality code with proper documentation was challenging.

  When talking about the rational for the deal, Snyk's CEO mentions that DeepCode will be '... constantly learning from the Snyk vulnerability database to become smarter'. 

Quick reminder: We mostly follow early stage startups that fits into three wide themes: data science tech stack, digital world/metaverse, and data science for the offline world. Therefore we may miss some exciting investments in 'traditional' SaaS startups and later stage deals.

Subscribe below for getting new reviews, video interviews, and info on events and meet-ups.