Skip to content

Startup

Application:

Memgraph 主打在数据流上做图计算 Katana Graph 图数据库与AI相结合 ArangoDB 多模态数据库

Katana Graph Capablities

GDB Applications

Batch

  • Identifying Essential Proteins with Betweenness Centrality in PPI network
  • Fraud Detection
  • Recommendation Engine

Streaming

  • Real-time Reddit analysis/Twitch Visualization

Recommendation System Startups?

General

  • https://www.4paradigm.com/product/intelligent_recommendation.html

  • https://www.recombee.com/

  • http://mr-dlib.org/

  • https://www.z5.ai/ai-powered-recommendation-as-a-service
  • https://www.webtunix.com/ai-powered-recommendation-as-a-service
  • https://kealabs.com/ (e-commerce)
  • https://yuspify.com/en/home/ (e-commerce)
  • https://www.bibtip.de/en (digital libraries)
  • https://exlibrisgroup.com/products/bx-recommender/ (digital libraries)

https://docs.google.com/spreadsheets/d/13Cd580vrysoQpcl2cgvD-99YUxiteb0v-ivzSdMPTUo/edit#gid=0

Vertical Frontier

  • Pandora Music
  • SmartNews
  • https://stack.g2.com/personalized-recommendations (Software)

How good is the idea of building a recommendation system along with a search engine as a startup?

https://www.quora.com/How-good-is-the-idea-of-building-a-recommendation-system-along-with-a-search-engine-as-a-startup

The idea is good, execution is tough. Let me elaborate.

The idea is good => finding relevant information that you ultimately want is a large problem. As information is piling up exponentially, discovering what you are really looking for becomes increasingly difficult. From my own experience we are building a recommendation engine for fashion products, a specific niche in the recommendation engine world. Discovering products that people want to buy is difficult because there is so much choice and the tough part is to filter through all the irrelevant products to discover the product that you actually want to buy. So the problem is there and it has not been solved well. When a person cannot find, say, a pair of jeans that she wants, it does not mean that this pair of jeans does not exist. It is more likely that she is not able to find that pair of jeans. Same could also be said about information discovery as well. Discovery of products / information is a large problem that has not been solved well and like with any large problems this also brings an opportunity. So you're on to something when working with recommendation systems and search engines.

The execution is tough => yet, whenever there is a large problem, there are many people solving it. Companies with more resources (budgets, brainpower) than your startup. So the trick is to outsmart them with your limited resources and the way you can do this is to pick a niche where you have a competitive edge. As an example, we're focusing on fashion product discovery and differentiating ourselves by building a visual recommendation engine. So pick a niche and differentiate. Whether your are combining certain type of data together that no one else is doing or applying a new algorithm or a new interface - it has to be something different and also something that you can execute at a niche level, to really be able to compete with larger companies.