Union.ai: Reliable and Scalable AI Orchestration

Ketan Umare, Co-founder & CEO

Built atop Kubernetes, Flyte is a cloud-native machine learning platform that uses computer languages, such as Python or Scala, to empower developers to build scalable data and ML workflows easily and efficiently. Designed and developed at Lyft, a ride-hailing app provider, the system is the brainchild of its former head of ML product engineering, Ketan Umare. Flyte enables the collaboration of machine learning engineers, data scientists, software engineers, product managers, and business managers as they work to provide value to their customers. The platform is currently in use at Lyft to efficiently use data to deliver accurate ETA and ride pricing details to customers.

Realizing that Flyte can benefit other industries that rely on machine learning products to scale, Umare, following his departure from Lyft, started his venture, Union.ai.

“People can easily observe, share and collaborate on a project, and reduce costs usually associated with machine learning operations,” says Umare, co-founder, and CEO at Union.ai.
Flyte makes it easy for all users to divide key tasks into reusable processes that remain effective even with new tech updates in a digital application workspace. Users can create, test, or verify models on a small Flyte local setup or a cloud-based Kubernetes environment. Flyte excels in transferring workflows and models from the desktop to a cluster. It allows machine learning engineers to smoothly run their workflows using their preferred tools like TensorFlow, PyTorch, Weights and Biases, Whylogs, BentoML, Dask, Ray, or Spark, in a highly collaborative Kubernetes environment.

Notably, some big names in the enterprise world, such as LinkedIn, Lyft, and Spotify, have recognized the operational value of Flyte's workflow orchestration system. For instance, at LinkedIn, developers use the zero-expense Flyte platform to create a model that can learn an individual's career choices and deliver relevant recommendations.

In addition to its open-source platform, Flyte also offers a SaaS platform called Union Cloud which is particularly useful for users or clients who would like to use Flyte, without the overhead of managing it. The company provides a fully managed Flyte service for clients, creating a secure and reusable ML environment that aligns with their businesses.
In a highly-regulated biotech and drug discovery space, Flyte's workflow orchestration system can ensure the scalability of next-generation sequencing operations or the repeatability of clinical test outcomes to demonstrate the efficacy of a medicine or ingredients. Clients can run multiple tests at a time rather than focusing on single variants and shorten periods for drug testing to minutes or hours compared to days. The streamlined workflow empowers users to reduce the complexity they usually face when producing FDA submissions in a validated environment. They can completely show the lineage of experimentation methods, the processes that worked, and other intricate details required to develop safe, reliable, and effective medicine.

People can easily observe, share and collaborate on a project, and reduce costs usually associated with machine learning operations

Since Flyte is open source, users can also benefit from a broader community. In addition, Union.ai also offers help, and staff members typically respond to questions within a few hours. In providing its workflow orchestration platform and services, Union.ai—a firm obsessed with customer success—hopes for users to always have the appropriate manner of learning, growing, and adopting a system within their company.