Union.ai is a software company that helps organizations build, deploy, and orchestrate large-scale artificial intelligence (AI) and machine-learning (ML) workflows and data pipelines. Its flagship platform is built around the open-source project Flyte, which is now deployed by many companies in production. Pulse 2.0 interviewed Union.ai co-founder and CEO Ketan Umare to gain a deeper understanding of the company.
Ketan Umare’s Background
Could you tell me more about your background? Umare said:
“My engineering background spans multiple industries, from banking to logistics, and I’ve focused on a range of topics, including cloud and AI platforms. I’ve had the opportunity to build initial versions of products at companies like Amazon and Citadel, and I was among the first few people working on Oracle Cloud Infrastructure (OCI) in Seattle. More recently, I led the machine learning (ML) team at Lyft, where we focused on using AI and ML to improve estimated arrival times (ETAs) and ride pricing for customers.”
“It was there, around 2016, that I discovered a gap between how ML models were being developed and how they were then being deployed into real-world applications. This realization ultimately inspired the creation of the open source project Flyte and, subsequently, the founding of Union.ai.”
Formation Of The Company
How did the idea for the company come together? Umare shared:
“We realized this had the potential to extend even beyond our open source project, Flyte, when we identified two core truths. First, AI and ML would become critical for businesses to compete, no matter the industry—even before the generative AI craze, we saw that AI would unlock efficiency gains that would become table stakes for most companies. Second, the traditional, linear approach to software development doesn’t apply to AI, which requires multiple iterations and experimentation, often in a short time period, and uses specialized hardware. These realizations inspired our team from Flyte to build the “new stack” for this next generation of AI and ML engineers, which is ultimately what Union.ai is intended to be. We want to make it easy to build AI products quickly, easily, and scalably.”
“Back in 2020, when we were first starting Union.ai, many companies were just beginning to grasp the unique challenges associated with building AI products. So, rather than trying to both educate the market and sell into it, we continued to invest in Flyte as an open source project. This gave ML engineers—especially those working at some of the companies that were at the forefront of AI innovation, who were beginning to understand the problem space—access to work directly with our technology in a collaborative way. As more companies realize the challenges of building and deploying AI, and as their needs scale, interest in a more robust, commercial solution like Union.ai has followed naturally.”
“Interestingly, as late as 2022, some investors I spoke to were still questioning the need for widespread AI integrations across industries—and then ChatGPT emerged. The rapid adoption of AI validated the strong belief that we’d been operating with: that AI held immense potential for businesses. That’s why we believe we’ve made the right investment to create a platform that empowers builders of future, AI-infused products.”
Favorite Memory
What has been your favorite memory working for the company so far? Umare reflected:
“One memorable moment happened early in Union.ai’s journey, in early 2022. We were working with a major, blue-chip pharmaceutical company that used AI as part of their imaging platform. After trying our product, an engineering leader at the company told me, ‘I’ve been an engineer for 15 years, and it’s rare that you get amazed by a product, and yet, that’s exactly what happened here.’”
“That feedback and validation of our work’s impact reinforced our determination to keep growing and strengthened our resolve to continue building.”
Core Products
What are the company’s core products and features? Umare explained:
“At its core, Union.ai is a developer suite designed to simplify the integration of AI into existing products and workflows. By managing infrastructure and orchestrating workflows, Union.ai accelerates iteration and streamlines production, supporting the entire AI lifecycle from model development and training, to real-time inferencing and deployment.”
“Three key challenges define AI development today: 1) a lack of tooling/platforms that adapt to the needs of everyone involved in the development cycle (like data scientists, ML engineers, and MLOps or DevOps practitioners); 2) runaway computing costs; 3) a slow time-to-market because of these inefficiencies. Union.ai addresses each of these challenges, providing an easy-to-use platform that meets the needs of everyone involved in AI development, while also including cost observability, security systems, and development guardrails to ensure compliance and cost management at the organization level—without slowing down development.”
“Several features make up our end-to-end platform. Union.ai can automate workflows and provide clear logs and records of activities, creating a single pane of glass into the AI development and deployment process. This makes it easier for multiple teams to collaborate. We also offer transparent cost management for computing by monitoring at the task level, enabling autoscaling, leveraging spot instances, and more. Finally, we help developer teams get to market faster with efficiencies like reusable containers, optimized parallelism, and full workflow caching. Our platform extends from model development and training all the way through real-time inferencing and deployment.”
Challenges Faced
Have you faced any challenges in your sector of work recently? Umare acknowledged:
“One of the biggest challenges is that many organizations have been slow to realize how, while every company is going to implement AI, the ones who will succeed will be those that can differentiate themselves through their unique data and internal AI systems. A year ago, for all the big talk in the market about being AI-first, most companies were just dipping a toe in the water—outsourcing an exploration of AI to a third party.”
“But now, just like we saw so long ago with traditional software strategy, there’s a pronounced shift happening from outsourcing AI development to owning your AI strategy in-house. Companies are seeing the opportunity in leveraging their proprietary data and expertise to build differentiated AI systems. At the same time, building an AI system is not easy. Companies need to become AI deployment machines, capable of rapidly taking ideas to production—but until now, the necessary infrastructure didn’t exist. Our team has built that foundational platform to bridge the gap.”
Evolution Of The Company’s Technology
How has the company’s technology evolved since its launch? Umare noted:
“The primary focus areas for AI development have shifted. Initially, model training, data processing, and hyperparameter optimization were central. While these remain important, with the advent of generative AI and the widespread availability of pretrained models, the focus has shifted to finetuning and, most notably, real-time inferencing.”
“At Union.ai, we continue to provide support for model training with features like activity logs and reusable tasks, but we’ve also adapted to add more support for working with generative AI models specifically. For example, as these models become ever-larger, deploying them is no longer like running a simple web application on Kubernetes. Today’s generative models require low-level optimizations and massive scalability to achieve predictable and acceptable performance. For example, if you were to ask Llama a question using traditional infrastructure, it might take upwards of 5 minutes to generate a response. Recognizing this performance issue, we’ve expanded our product at Union.ai to include a specialized infrastructure capable of delivering Llama responses in under a minute, even from a cold start.”
Significant Milestones
What have been some of the company’s most significant milestones? Umare cited:
“Some of our favorite milestones are when we reach new heights of helping people develop and deploy AI more efficiently! For example, over 3,000 teams are now leveraging Union.ai’s technology to power their AI initiatives, including more than half of the “Magnificent Seven” technology leaders.”
“We also recently introduced real-time inferencing support, enabling customers to deploy and serve their models in production at the speed of business operations.”
Customer Success Stories
When asking Umare about customer success stories, he highlighted:
“Customers turn to us when they need an innovation partner to help them build a scalable AI/ML strategy. At Woven by Toyota, Union.ai scaled autonomous vehicle development pipelines, accelerating machine learning iteration cycles by over 20x and saving millions annually in infrastructure costs. Rezo, a biotech pioneer, leveraged Union.ai to dramatically scale their protein prediction workflows, achieving up to 90% cost reductions and enabling rapid advancements in drug discovery, without added DevOps overhead. Additionally, Spotify streamlined complex financial forecasting across multiple teams, reducing quarterly forecasting cycles from four weeks to under two, while significantly boosting accuracy and operational agility.”
Revenue
When asking Umare about the company’s revenue details, he revealed:
“Building on a strong 2025, where H1 marked our largest phase of growth so far, we’re excited to carry this momentum into the new year.”
Total Addressable Market
What total addressable market (TAM) size is the company pursuing? Umare assessed:
“Looking ahead, we see a clear reality unfolding: within the next five years, every Fortune 2000 enterprise and every tech company is going to need a meaningful AI/ML strategy to stay competitive. Union.ai is designed to enable that—providing agile, developer-friendly, and cost-effective infrastructure so teams can confidently build and scale their AI initiatives. We’re positioned exactly where the market needs us, and that’s driving tremendous momentum.”
Differentiation From The Competition
What differentiates the company from its competition? Umare affirmed:
“Union.ai is what Amazon SageMaker and Bedrock would be if they were built for developers to love, not for locking customers into cloud contracts. We deliver more optimized, enterprise-grade orchestration and AI infrastructure—but with real openness, extensibility, and developer-first design. Teams using Union.ai can scale massive, compound AI systems across any environment without burning through their compute budget or locking into an inflexible ecosystem. We believe AI infrastructure should be empowering, not restrictive, and our deep commitment to open source through Flyte proves it. Union.ai is the platform you choose when you actually want to move fast, save money, and build the future on your terms.”
Future Company Goals
What are some of the company’s future goals? Umare concluded:
“We’re excited to make development of compound AI systems—systems leveraging an ensemble of models and data—as easy as traditional software development. That means continuing to improve our real-time serving, best-in-class workflow orchestration, and infrastructure integrations.”
“Central to our philosophy is the belief that open source is foundational to innovation and trust in AI; that’s why we’re deeply committed to continually investing in Flyte and fostering an open, collaborative community around it. We also plan to strengthen our position in verticals such as autonomous systems, biotechnology, and financial services, enabling technical teams everywhere to accelerate innovation, optimize resources, and scale compound AI systems effectively.”
