Neuromation Bi-annual Report 2H 2020

neuromation report 2020

Introduction

As discussed in our last update, these voluntary reports will now be on a bi-annual cadence, so the below will cover from June 30 to December 31, 2020 as well as some more current information that may have become available prior to publication.

We very much appreciate the opportunity to update our token holders on the latest events at Neuromation, including product development, research and client work. These reports also provide an excellent opportunity to review our available services for business clients and token holders as offered by our “Neuromation Labs” business practice units.

  • As our main focus continues to be the development of our MLOps solution, Neu.ro, this latest update will include significant detail on new feature additions and progress there as well.
  • Importantly, we would also like to announce and call attention to our newly launched website https://neu.ro/, which includes expert content from our advisors and senior executives, and insights on the application of AI across various industries.
  • Finally, we would like to share some new client news as well as list various exciting new partnerships that Neuromation is involved in.

Product Update

At Neuromation, our primary hypothesis is that the lack of robust, adaptable Machine Learning Operations (MLOps) is a key obstacle to the scaling and widespread adoption of Machine Learning.

Readers of these reports as well as our customers will know Neu.ro as a complete MLOps Solution combining hybrid resource orchestration for ML workloads with fully customizable pipeline creation and management, integrated collaboration tools, automated instance control, enterprise features and full support from our remote MLOps team.

But in a field as young and dynamic as machine learning, we find it essential to regularly make fine-tuned adjustments to our strategy while always remaining true to our core mission. In this regard, we benefit greatly from our direct market experience facilitating AI Transformation efforts for corporate customers in a variety of industries.

Over the past six months, client experience has repeatedly demonstrated to us that virtually all AI development teams, be they at large enterprises or new startups, begin by spending the first 3–6 months building their first ML pipelines from available tools. These custom integrations are time-consuming and expensive to produce, can be fragile, and frequently require drastic changes as project requirements evolve. Frequently, these custom ML pipelines only support a small set of built-in algorithms, or a single ML library and are tied to each company’s infrastructure. Internal users in these environments cannot easily leverage new ML libraries or share their work with a wider community.

As a result of these learnings, we are convinced that the best path to robust, adaptable Machine Learning Operations (MLOps) is to maintain agility and avoid the pitfalls of technical debt by facilitating the seamless connection of the ever-expanding universe of ML tools into your workflow. In this way, we cover the entire ML lifecycle from Data Collection to Testing and Interpretation.

Our product team over the past six months has focused on implementing robust out-of-the-box integrations between Neu.ro and best-in-class tools for each stage of the machine learning development cycle. Using Neu.ro, the various components of a machine learning workflow can now be treated as independent, reusable, modular parts that can be pipelined together to create, test and deploy models.

In order to account for the wide range of use cases, environments, and user preferences, Neu.ro now provides integration with various open-source and proprietary tools for each of the following steps in the MLOps process:

mlops process

All resources, processes and permissions are of course managed through our Neu.ro platform and can be installed and run on virtually any compute infrastructure.

Neuro platform has been adapted to run in cloud-only, on-premise and hybrid cloud configurations in order to serve increasing business demand for AI/ML development and hosting services via regional cloud providers.

Neu.ro Features Update

Since our last update, there have been many new features added to our MLOps solution Neu.ro, as well as some important changes under the hood that improve efficiency and stability.

In terms of cost and resource management, Neu.ro now boasts improved scheduling capabilities, which allow organizations to share limited computational resources in a fair way. We have also added a way for managers to limit the amount of simultaneously running jobs per user.

Improving our on-premise capabilities, Neu.ro now supports persistent disks (fast storage that can be used for the temporary storage of the large datasets and artifacts).

Neu.ro also now provides more flexibility in terms of installation types, allowing for self-contained offline installations and for all platform cluster components to be executed on a single machine if necessary.

Our custom workflow and pipelining engine, called Neu.ro Flow, has reached sufficient utility and stability that it has now become the main entry point for any project.

In the time since we last reported on progress, Neu.ro Flow has added several new features, such as batch-mode task caching, to help speed up the development process and optimize resource consumption throughout the project life cycle.

Actions in Neu.ro Flow are added as reusable components (with a dedicated GitHub organization at https://github.com/neuro-actions ). All recipes (such as this and this) and project templates are also now based on Neu.ro Flow.

Finally, we have made numerous UX improvements, such as reworked job details in the web UI and managing presets in both CLI and web UI, the `neuro top` feature for running multiple jobs, and others.

Company Update

As Neuromation ML services and the Neu.ro MLOps platform begin to gain broader recognition, we are proud to announce that Neuromation is now listed as one of the most popular “MLOps services” in the world with the top position on the first page of Google search results. Neu.ro is also now in the first page of results for the term “MLOps Platform”. We have also reached #13 overall for the “MLOps” category in search results and are climbing. While these new categories are in the process of formation, it is very important to be part of the landscape and we are glad to announce that our contribution to the core technological standards of MLOps is being recognized by key industry players, the professional community, and ML practitioners alike.

Neuromation is also pleased to announce that we concluded several high-level partnerships during the period.

  • Neuromation with Neu.ro MLOps platform is now a Core member of the AIIA (AI Infrastructure Alliance). The AIIA is seeking to facilitate the evolution of a Canonical Stack (CS) for machine learning through the input of other core members including Determined AI, Pachyderm, Seldon, Clear ML, Tecton, New Relic, Arize, Valohai, YData, Algorithmia, Fiddler and of course Neu.ro. At the time of publishing, the AIIA includes over 30 member organizations working in the ML space.
  • The company also recently became a member of CAI (Content Authenticity Initiative) alongside Adobe, Microsoft, The New York Times, Twitter, Qualcomm, BBC, and more. CAI aims to design an open and extensible approach for providing media transparency that allows for better evaluation of content provenance.
  • Neuromation is also an AWS Partner, Microsoft Partner, Google Cloud Partner, and a member of Microsoft Startups and the NVIDIA Inception program.

As an additional way to increase utilization of NTK, Neuromation has begun to test deployments of a payment gateway for general AI/ML services and computing power in pilot projects with regional cloud providers in EMEA. Given the growing interest in MLOps solutions and services from data centers, driven by increasing demand for Deep Learning computation, we see this as a potential channel for accelerating utilization of NTK.

Finally, we would once again like to extend our invitation to all interested customers and token holders looking for solutions to specific AI-driven problems to get in touch — our AI services marketplace for token holders remains active. Neurotoken (NTK) can still be exchanged for AI services including data labeling, model development, synthetic data creation, AI strategy consulting, and fundamental research. Engagements with any of our Neuromation Labs units (MLOps Lab, Synthetic Data Lab, Animal AI Lab) can also be transacted through the Marketplace in exchange for NTK. Please contact us for access and pricing assistance.