← All organizations

Open-source AI inference infrastructure

Superlinked

Superlinked builds Superlinked Inference Engine (SIE), an open-source platform for self-hosted inference used in AI agents, search and document processing. Developers can combine embeddings, reranking, OCR, structured extraction and language models behind one API. SIE runs in customers’ clouds or air-gapped environments, allowing teams to retrieve context, convert documents into Markdown and execute agent workflows within their own infrastructure.

Founded in 2021 by CEO Daniel Svonava, formerly a YouTube tech lead, and COO Ben Gutkovich, formerly a McKinsey strategist, Superlinked initially developed vector-search infrastructure. Its framework combined structured metadata with unstructured content in configurable multimodal embeddings for recommendations, retrieval and analytics. In 2026, the company deprecated and archived that framework in favor of SIE. The inference engine packs multiple specialized models onto shared GPUs; a cluster-wide queue lets workers pull requests and assemble batches.

By August 2026, Superlinked reported more than $12 million in cumulative funding, naming Index Ventures, Theory Ventures and Samsung Next among its investors. Its earlier $9.5 million seed announcement in 2024 also marked a commercial milestone: the company reported closing its first six-figure customer contract in 2023. Self-hosting is free, while managed hosting and an agent plugin are listed as upcoming.

superlinked.com

1 talk

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Messages from the stage

Sharing GPUs across models

Makraduli describes loading and hot-swapping multiple models on shared GPUs to improve utilization. The talk also examines token-padding inefficiencies as a serving concern.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-28