Derwen, Inc.

For AI apps, what’s the gap between research claims in the latest papers -- and having independently verified code other people can use in production?

Currently in machine learning, only a small percentage of research papers can be verified by independent review of their code. Many published results cannot be reproduced.

Meanwhile in enterprise organizations, options are limited. Your data is too sensistive to use with SaaS vendors. Paying a system integrator to build internal tools is iffy: given how fast technology is evolving, whatever you spec will likely become outdated before you can use it much.

Your staff might build internally with open source, but do they have the time and experience to evaluate the torrent of open source projects coming out in conjunction with latest research? For that matter, which projects fit your use cases and would be viable for apps in production?

This is where Derwen https://derwen.ai/ plays a vital role. Our team bridges the gap for your team. We’re experts at working with open source developer community, experts at understanding how to evaluate libraries, fitting available technologies to your use cases. Experts at finding these kinds of answers from within communities of practice, quickly. Experts at explaining these technologies and their trade-offs.

• /graph resource pages
• blog, newsletter
• industry analysis
• vendor-neutral listings

• textgraphs
• pytextrank
• kglab
• disparity_filter
• mkrefs
• "Graph Levels of Detail"

• natural language applications, especially extracting semantic info markup from unstructured data sources, for building graphs

• graph technologies, bridging between the camps of graph DB, graph algo, graph ML, etc.

• open source integration, including packaging, performance evaluation, concurrency/parallelization, and preparation for independent security audits

• "last-mile" integration into operations research, optimization, recsys, etc.

• enterprise applications, particularly in highly regulated environments