About Us

Our mission:

Rapidly innovate in photonics and connectivity to enable the massive scaling of AI infrastructure.

Our story

At Lumilens, our team has spent decades designing the optical and silicon platforms that defined how data moved across the cloud and the Internet. But as AI workloads have exploded, electrical and legacy optical interconnects have reached their limits.

So, we set out to build a new, photonics-based foundation for AI infrastructure and data centers. For both scale out and scale up architectures.

But designing next-gen AI connectivity is not enough – you have to build it...smart. Lumilens has invested in cutting-edge, custom robotics and AI-optimized production processes.

VP/GM India & founder

VP Products & founder

VP Operations & founder

Harish Devanagondi

VP Engineering, Silicon

VP Manufacturing & Ops

VP Silicon Photonics

VP Silicon Photonics

Launch press release

AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires:AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires:

Launch press release

AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires:AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires:

Launch press release

AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires:AI workloads differ from conventional cloud applications. Instead of relying on independent compute nodes, AI training clusters must act as a single, tightly coupled supercomputer. This requires: