Artificial orchestrates labs and augments scientific research with Azure and Azure AI services

Bringing automation and AI to life sciences labs

Discovering and creating new drugs takes the coordinated efforts of hundreds of scientists, operators, and technicians, often across multiple sites and countries within different pockets of an organization. As a result, scientific labs are notoriously complex environments with massive datasets, which makes maintaining efficient workflows and data accuracy challenging. Manual processes and scattered data can lead to errors, delays, and increased operational costs.

Given the obstacles, life sciences professionals face immense pressure to run their labs with precision and speed. Artificial helps them do both using natural language processing and AI. The startup’s software suite, the Artificial Platform, helps life sciences laboratories digitize their scientific work and get data off their instruments and equipment more efficiently. “We’re tool builders who like making elegant, stylish tools for scientists and other people to accomplish tasks they couldn’t normally do on their own,” says David Fuller, Chief Executive Officer and Cofounder at Artificial. 

The goal is to make lab experiments reliable, repeatable, and trusted. Artificial’s founders knew this was a great opportunity for the cloud. “Using the cloud gives you global visibility and scale, particularly along the data storage front but also with real-time observability of what’s happening as different scientists manage different parts of the process,” says Fuller. 

Artificial also wanted to unlock the future for life sciences professionals through AI-driven labs. This led to the creation of a cloud-to-edge connected productivity layer for labs to conduct more efficient experiments—spanning their researchers, equipment, and locations—and achieve better, faster results. “We’ve taken a legacy control system and reimagined it as a born-in-the-cloud control system, globally distributed across cloud infrastructure and edge devices,” says Fuller. “It’s a containerized set of microservices that runs on Kubernetes and delivers live, globally centralized visibility into experiments.”

Building around Azure Kubernetes Service, Azure OpenAI Service, and digital twins

To gain a global platform that’s ready to scale its AI solution, Artificial migrated from its previous cloud platform to Microsoft Azure, taking advantage of Microsoft APIs and working with Microsoft Global Black Belts to provision and commission applications and infrastructure tailored to life sciences for the cloud. “Microsoft is a trusted brand for security and data governance, with built-in tools and building blocks that help us stay compliant in accordance with this industry’s best practices,” says Fuller. “For example, we have to be ISO certified, and moving to Azure helped us achieve that.”

Artificial’s system architecture now comprises a static front end with a dynamic, responsive web app built using Vue.js. A fleet of services behind it is orchestrated via Azure Kubernetes Service (AKS) and sent to Azure Virtual Machines running Vanilla OS, a Linux on Azure–based operating system. “Many of our services are internally stateless but need some way to offload stateful queries and process events, and we use a combination of Redis and PostgreSQL for that,” says Charles Crain, Chief Technology Officer at Artificial. “We store every step of every job in a queryable PostgreSQL database, and we use Redis as our message broker.” All the cloud applications, plus point-in-time backups for PostgreSQL and Redis, use Azure Blob Storage.

As Crain explains, “Our product is based around a workflow that runs in a laboratory authored using a domain-specific dialect of the Python programming language, Orchestration Python, that’s designed to pair with a language model–based assistant. But you can’t just ask any off-the-shelf model to work with Orchestration Python or to have contextual awareness of a customer’s lab.” In response, Artificial developed a prototype bot based on Azure OpenAI Service and Azure AI Search. “Customers will have greater success with our product if we augment it with context,” adds Crain. “Our Azure-based bot uses retrieval-augmented generation (RAG) to guide customers through coding in our specific Python dialect to work in their particular application domains.”

Artificial’s Digital Twin provides customers with a 3D digital twin of their laboratories, offering both a spatial picture of the lab and rich experiment observability through an API. “You can view state changes in your lab in real time via your browser and see bits of the digital twin highlighted as the workflow starts activating equipment in the lab,” notes Crain. “That’s all events flying back and forth over Redis, so Redis and the digital twin come together to form the full data picture of the lab. We also plan to help customers fine-tune their workflow designs by bringing in context for the data model that gives the coordinates of a piece of equipment or how many pieces of equipment are in those positions.”

Artificial is now using off-the-shelf technologies such as managed AI or semantic search from Azure to deliver rapid business benefits. As they manifest, scientists step into self-driven labs where they’re now able to elevate their proposals for further drug experiments and automate overhead to be involved in design from a more strategic and truly experimental capacity versus completely hands-on. In the case of drug discovery and molecule design, pretrained AI models can suggest new experimental candidates or "score" experimental candidates, which helps scientists avoid excessive and expensive wet lab time and vastly increases the number of candidate molecules that can be screened. More molecules screened more quickly at a lower cost per data point allows scientists to arrive with more effective drug candidates and bring them to market more quickly than was previously possible. 

Supporting industry breakthroughs

Since making AI more accessible in lab settings, Artificial has made tangible impacts on the industry. Using the Artificial Platform, a leading US pharmaceutical and biotechnology company was able to fast-track how it creates vaccine delivery mechanisms for its therapeutics. To support other areas of the business’s development of therapeutics, Artificial created libraries of different lipid nanoparticle properties and characteristics, which frees up time for scientists to conduct additional research and create more particles. Artificial is also helping a cell and gene therapy company, which focuses on novel therapies through gene editing, use the Artificial Platform to gather more information ahead of development discussions. Its staff is then able to use automated lab equipment more optimally with fewer errors and more accessible data and unlock insights to continuously improve time and cost per data point.

Artificial’s pharmaceutical customers, who must safeguard their information, share that they’re more comfortable coming on board because Artificial runs on Azure alongside Azure connectors. “The ability to have Azure instances connected on-premises inside the bounds of the pharma companies is going to be of major value for the truly fundamental, critical data that they simply won't let out of the lab,” says Fuller. “We get and provide all the benefits of Azure without any trust or security concerns because they're very comfortable that it’s all within their control.”

“Customers use our platform and tools to accelerate drug discovery,” adds Fuller. “It makes me proud to help these world-class, brilliant minds solve problems that are fundamentally relevant to our lives—and to have a team and a set of solutions that they like. I love making our customers successful with a group of people I like working with.” 

“Making more drugs available to more people in more ways”

By digitizing lab operations, Artificial helps its customers accelerate training, reduce errors, and enhance the overall efficiency and reliability of scientific research. “There’s this sweet spot in the lab for technology that’s substantially useful, optimal from a cost-savings point of view, more effective, more repeatable, more seamless, easy, and approachable with a low barrier to entry so that more people can adopt it,” says Fuller. “The ubiquitousness of the cloud makes us better able to deliver the digital experience through whatever device scientists and staff want to use.”

Artificial is at the cusp of building the sustainable architecture and customer-facing features that would automatically give life sciences customers an enhanced, fully semantics-capable copilot they can query about their lab and everything in it. “The data that’s produced all comes through Azure AI Search, RAGing, and Azure OpenAI language models via tools we’ve written to enable agency for the lab,” says Fuller. “AI is just a brain in a jar if it doesn’t have something to tell it what to do or to ask about, so we want to be that interface. And because we’re already running the experiments and gathering the data, we obviously want to be what collects the data and standardizes its shape.”

Artificial is rapidly tracking growth on multiple axes—more customers, bigger customers, more GPU supercomputing, and bigger customer labs with more data points and more experiments in parallel. “The world is making more drugs available to more people in more ways, and we’re helping enable that across the markets,” says Fuller.

Discover more about Artificial on LinkedIn and YouTube.