Engramme seeks major funding to build “AI that never forgets”
Engramme, a memory-focused artificial intelligence startup founded by Harvard neuroscientist Gabriel Kreiman, is seeking to raise about $100 million at a valuation that could reach $1 billion, according to Bloomberg. The terms are not final, and investor discussions are ongoing.
The company was founded in 2025 by Kreiman and co-founder Spandan Madan. It emerged from stealth last month after previously raising a $3 million pre-seed round led by Mayfield Fund alongside other Silicon Valley investors.
From neuroscience research to commercial AI memory
Kreiman is known for early work in memory neuroscience, including a 2000 Nature paper co-authored with Christof Koch at Caltech that explored how individual neurons respond when people see images and when they imagine them. After decades in academia—including roles at Harvard Medical School and Boston Children’s Hospital—he left to commercialize his research on how the brain encodes memory.
“Large Memory Models” and the enterprise angle
Engramme says it is developing Large Memory Models, positioning them as a new class of AI systems aimed at storing and retrieving information more like the human brain than traditional databases or search indexes. The company highlights three core properties: lifelong storage at petabyte scale, proactive retrieval that surfaces relevant information without a query, and associative recall that links information across time and context.
A consumer iOS app is in beta, while an enterprise API for memory extraction and retrieval is under development. The company says conversations with more than 50 potential users surfaced demand ranging from consumers managing memory loss to organizations seeking to preserve institutional knowledge when employees leave.
Competition and investor questions
The space is increasingly crowded, with competitors including Mem0, Rewind AI, Zep, LangMem, and MemGPT. Engramme argues its neuroscience-first approach could outperform systems built around vector databases and embeddings, but independent benchmarks have not yet been published. If completed, the round would represent an aggressive step-up from a pre-seed stage, putting pressure on the company to prove technical differentiation and product-market fit.
Related
