Introducing the first of the blog series: Powered by Lemma. Learn More
October 1, 2024
But our journey started long before this moment.
Back when we were at Palantir, we witnessed hundreds of different AI use cases across Fortune 100 companies and government agencies building AI systems in some of the most stringent environments. In doing so, we saw a clear pattern emerge: the most successful enterprise applications of AI, especially those used for workflow automation, were deeply integrated across teams and embedded into every part of the business.
We noticed that decision-making in enterprises is often federated across people and business units. A data scientist isn’t going to choose a Salesforce license, just like an engineer isn’t picking the AI model, and an operations analyst isn’t deciding on database storage. Each of these roles contributes to a larger workflow, but there wasn’t infrastructure designed to support this kind of cross-functional collaboration.
That’s when the idea for Thread AI and Lemma was born.
We built Lemma to solve this exact problem: creating an AI workflow infrastructure that’s composable, reusable, and modular. The name “Lemma” is inspired by mathematics, where a lemma is a “helping theorem,” an incremental step toward solving a larger proof. That’s what Lemma is for AI—an infrastructure made up of foundational building blocks, designed for workflows that can be adapted and scaled without breaking down.
Finding the right software to support these kinds of workflows is hard. It is even harder if there is no in-house expertise to build something yourself. In addition, the broader AI tech stack has been evolving rapidly, with new technologies and derivatives of the same popping up rapidly over the past few years.
AI models and data come in different shapes and sizes.
There are no unified standards around data models for AI services. By design, Classical and Generative AI models are built to operate on all kinds of data, structured and unstructured alike. AI services require a data translation layer.
Good AI should be trustworthy, but often needs human augmentation.
Advancements in AI have opened up endless possibilities for innovation, but AI workflows are iterative and can require guidance and direction, despite even rigorous evaluation. The proper human oversight and guardrails must exist at critical stages, to incorporate contextual understanding and subjective evaluation.
Most enterprises use hundreds of software products, some legacy, some cutting edge.
Purpose-built infrastructure is needed in order to connect and orchestrate AI across different products, platforms, and cloud providers, as well as to handle different error codes, authentication policies, and API protocols. Organizations want to track the entire life cycle of their operations from start to finish and want to know how their data is used and where it’s going.
The modern enterprise has invested millions of dollars and thousands of human-capital hours building infrastructure, processes, and procedures around critical business operations. At the same time, AI is shaping the future of industries everywhere. Fortunately, leveraging AI does not have to require a complete organizational overhaul.
The best kinds of technologies are transformational, but transformation doesn’t need to be at the cost of safety and intentionality. True, systemic transformation takes time - even when technology replaces whole teams or infrastructures - new supporting processes, educations, integrations, and markets arise. The system has to acclimate and catch up.
Whether it’s ensuring your guests have the best personalized experience at your luxury hotel or equipping your field technicians with the right information faster for machine maintenance, we fundamentally believe that every enterprise workflow is a mission critical workflow and should be treated as such.
The recent and massive boom in commoditization of both proprietary and open source AI models and services have made AI more accessible than ever. Coupled with the ubiquity of infrastructure and open source advancements, there is no need to compromise on scale, durability, or observability, regardless of the industry in which you operate.
We aim to democratize mission critical software.
Angela McNeal & Mayada Gonimah
Compliance
CJIS
GDPR
HIPAA
SOC 2 Type 2 - In progress