Lamini builds a generative AI platform purpose-built for enterprise deployments, emphasizing accuracy, scalability, and security. The company says its stack is optimized for enterprise workloads across hardware and software, and it touts a technique called "memory tuning" to reduce hallucinations by training models to recall exact proprietary facts. Lamini supports on-premises, public and private cloud deployments, including air-gapped environments, and claims it can scale elastically to over 1,000 GPUs. The company runs many models on AMD Instinct GPUs and says AMD supplied early data center hardware. Lamini has a roughly 10-person team and counts AMD, AngelList and NordicTrack among early paying users, along with several undisclosed government agencies. The startup reports it has raised $25 million across seed and Series A and is using the funds to triple headcount, expand compute infrastructure, and pursue deeper technical optimizations; leadership says its gross margins and burn look like a regular tech company.

Total raised

$25M

Funding rounds

1

Latest round

Equity

Latest activity

May 2024