Engram has launched from stealth with $98 million in funding to solve a costly problem plaguing enterprises: AI systems that must relearn the same organizational context on every single query, wasting computational resources and money. The Stanford-founded startup is building what it calls a "learned memory layer" that trains AI models to study an organization's unique knowledge, documents, and workflows in advance, compressing that information into compact, reusable memory that improves continuously with use.

Why Does AI Keep Forgetting Your Organization?

Today's large language models (LLMs), the AI systems powering tools like ChatGPT and enterprise assistants, operate like brilliant strangers. They can synthesize vast amounts of information and solve complex problems, but they know nothing about your company's specific processes, culture, or institutional knowledge. Every time an employee asks a question, the AI rereads the same documents, relearns the same context, and rediscovers the same information from scratch. As enterprises deploy AI agents across every function, those wasted computational tokens are becoming one of the biggest hidden financial pressures in enterprise technology.

Engram's approach flips this model. Instead of pasting organizational knowledge into conversations on the fly, the company trains models to anticipate an organization's questions and prepare memories ahead of time. The result is AI that gets smarter the longer it's used while consuming up to 100 times fewer tokens than traditional approaches.

"Whatever the AI knows about you is improvised on the spot, a sticky note about your past, a document pulled mid-conversation. If we can anticipate your interactions, we can prepare memories ahead of time instead of pasting them on the fly," said Dan Biderman, CEO and co-founder of Engram.

Dan Biderman, CEO and co-founder of Engram

How Does Engram's Memory Layer Actually Work?

Engram's founding team brings deep academic expertise in how AI learns and remembers. CEO Dan Biderman completed postdoctoral work at Stanford under Chris Ré, one of the most influential figures in modern machine learning, where he focused on making AI agents cheaper to run. His co-founders span leading research labs working on AI learning and memory.

The technical challenge is significant. When an AI reads a 70,000-word legal contract, roughly 400 kilobytes of text, its internal memory of that document can swell past 100 gigabytes, making the model slow and expensive to run. Engram's approach compresses that learning into compact memory that the model can reuse on every query, doing the heavy lifting once rather than repeatedly.

The company's founding team includes researchers who developed canonical methods for this work:

• Sabri Eyuboglu, CTO: A Stanford PhD under Chris Ré who created Cartridges, a method for turning large bodies of documents into small, reusable memory structures.
• Jessy Lin: A Berkeley PhD who researched at Meta's FAIR lab and developed Active Reading, a technique for training models to study material deeply rather than simply store it.
• Jack Morris: A Cornell PhD from Meta's FAIR lab known for work on retrieval and memorization in large language models.
• Scott Linderman: A tenured Stanford professor of statistics and neuroscience researching state space models, a fast-growing alternative to transformer architectures designed to handle long stretches of information efficiently.

Who's Backing This and Why Does It Matter?

Engram's $98 million Series A funding round was led by General Catalyst, Kleiner Perkins, and Sequoia Capital, with participation from Factory, Modern, Amplify Partners, and Neo. The round also includes backing from notable angels including Andrej Karpathy, co-founder of OpenAI, and Assaf Rappaport, co-founder and CEO of Wiz, signaling confidence in the approach from leading AI figures.

The company has already secured meaningful commercial partnerships. Microsoft is testing Engram's memory layer within Microsoft 365, with a commitment to GPU capacity across Dapple and Azure to train models at scale. Notion is integrating Engram's models into its new custom agents, and Harvey, a legal AI platform, is working with Engram to build learned enterprise memories that are secure and cost-efficient.

"Our customers have built up extraordinary knowledge inside Microsoft 365, and we've only begun to tap what it can do for them. Engram's approach could turn that knowledge into a kind of memory each organization owns and controls, while making AI efficient enough to power the long-running, proactive agents we believe every knowledge worker will eventually rely on," said Jason Graefe, Corporate Vice President of AI Partner Catalyst at Microsoft.

Jason Graefe, Corporate Vice President of AI Partner Catalyst at Microsoft

What Makes This Different From Other Enterprise AI Solutions?

In today's AI landscape, model providers accumulate the value generated by every enterprise interaction. Engram offers a fundamentally different model where companies own the intelligence they build. The more an organization uses Engram, the more specialized and proprietary its models become, creating a form of AI that is sovereign to the enterprise and not dependent on or extractable by any model provider.

This addresses a critical pain point for enterprises deploying AI at scale. Most conversations around enterprise AI have focused on making models generally smarter, but for companies actually running AI in production, that was never the hard problem. Getting a model to truly remember a specific organization and its unique ways of working is the challenge that has lacked a convincing solution until now.

"Memory is the missing ingredient in AI. We see enormous potential for Engram's technology across the companies we're building and transforming in healthcare, legal, and financial services, where the institutional knowledge is deep and the cost of running AI against it is only growing," said Hemant Taneja, CEO of General Catalyst.

Hemant Taneja, CEO of General Catalyst

Engram's emergence reflects a broader shift in enterprise AI investment. Rather than chasing general-purpose model improvements, venture capital is increasingly flowing toward infrastructure and tools that solve specific, costly problems for organizations already deploying AI. The company's funding and partnerships suggest that the market recognizes organizational memory as a critical missing piece in making enterprise AI both effective and economically sustainable.