Harvard neuroscientist eyes $100M raise at a $1B valuation to build AI that never forgets — TFN

In 2000, Gabriel Kreiman revealed a Nature paper exhibiting that single neurons within the human mind hearth each when an individual sees a picture and once they think about it. Co-authored with Christof Koch at Caltech, this was a landmark in reminiscence neuroscience.

Twenty-six years later, Kreiman, then a professor at Harvard Medical Faculty and Boston Kids’s Hospital and affiliate director of the MIT-Harvard Centre for Brains, Minds and Machines, left academia to commercialise his analysis on how the mind encodes reminiscence.

Kreiman’s startup, Engramme, based in 2025 with co-founder Spandan Madan, emerged from stealth final month and is in search of to raise about $100 million, in accordance with Bloomberg. Traders have mentioned a possible valuation as much as $1 billion, although phrases usually are not remaining.

Beforehand, the corporate raised a $3 million pre-seed spherical led by Mayfield Fund and different Silicon Valley traders.

Engramme is creating Giant Reminiscence Fashions, a brand new class of AI techniques designed to retailer and retrieve info extra just like the human mind does than conventional databases or search indexes do.

The structure focuses on three key properties: lifelong storage at petabyte scale, proactive retrieval that surfaces related info and not using a search question, and associative recall that connects info throughout time and context, just like how the hippocampus hyperlinks associated recollections.

The corporate’s client iOS app is in beta, and an enterprise API for reminiscence extraction and retrieval is in growth. By means of discussions with over 50 potential customers, together with older adults with reminiscence loss, mission managers, and AI builders, the workforce recognized each a client market and a transparent enterprise use case: preserving organisational data typically misplaced when workers depart.

Direct opponents embody Mem0, Rewind AI, Zep, LangMem, and MemGPT. Engramme claims its method is grounded in neuroscience moderately than engineering conference, arguing that constructing reminiscence techniques based mostly on how the mind works yields essentially totally different and higher outcomes than utilizing vector databases and embedding fashions. It is a testable declare, however the firm has not but supplied impartial benchmarks.

If accomplished at a $1 billion valuation, the $100 million elevate can be an aggressive pre-product entry for a corporation with solely a $3 million pre-seed and a beta app. Traders should determine whether or not Kreiman’s 20 years of peer-reviewed reminiscence analysis present a real aggressive benefit or only a compelling narrative.