Cerulion Inc.
Cerulion builds an open-source, ROS 2-compatible robotics runtime and harness that gives AI agents a safe physical interface to real robots. Its tooling is designed to make robot communication faster and more reliable, while helping teams record, investigate, replay, and verify robot runs.
At a Glance
- Robotics startups and product teams
- Embodied-AI developers
- Industrial and research robotics teams
- Teams using ROS 2
- +2 more
AI Tools by Cerulion Inc.
(1)Cerulion
Open Source Robotics Runtime
Discussions
No discussions yet
Be the first to start a discussion about Cerulion Inc.
Latest News
Published engineering note: How ROS 2's DDS layer actually works.
Cerulion demos at Circuit Launch's November 2025 robot block party.
Cerulion makes The Robot Report's 100 Robotics Startups to Watch 2025 / Startup Radar.
Cerulion wins the 2025 Pitchfire Startup Competition at RoboBusiness.
Products & Services
Open-source Rust runtime and command-line tooling for composing robot applications as node graphs communicating over zero-copy shared memory; licensed AGPL-3.0-only, with a commercial license available.
Visual investigation interface for robot runs, including synchronized camera views, topic lists, and joint plots.
Rolling in-memory robot-history capture that preserves recent history as an MCAP clip on demand or when a runtime fault occurs.
Early-access analysis product that finds the moment a robot run went wrong and cites claims to topics and timestamps in an MCAP recording.
Market Position
Cerulion positions itself as a high-performance, developer-friendly alternative or complement to conventional ROS 2/DDS middleware, emphasizing zero-copy transport, fault isolation, recording/replay, and AI-agent-assisted investigation. Crunchbase identifies NVIDIA, ANYbotics, and PickNik Robotics as possible alternatives or competitors; Cerulion differentiates through its open-source Rust runtime, native MCAP verification workflow, and controlled MCP interface for AI agents.
Leadership
Founders
Lakshay Sharma
Cofounder and CEO. Studied computer engineering, astrophysics, and robotics at the University of Pennsylvania and researched robot motion planners at MIT; he is described by YC as an MIT robotics PhD founder.
Se Hwan Jeon
Original cofounder. An MIT robotics PhD student who worked with Lakshay Sharma on robotics middleware and robot reliability.
Raveen Kariyawasam
Current founder/team member shown by Cerulion as working with Lakshay on founder sessions. His background includes a Rhodes Scholarship, a PhD in engineering from Oxford, bioengineering study at the University of Pennsylvania, and founding/leading Simplifine (YC S24) before joining Cerulion.
Executive Team
Lakshay Sharma
Cofounder and CEO
University of Pennsylvania computer engineering, astrophysics, and robotics background; MIT robot-motion-planning researcher and robotics PhD founder.
Raveen Kariyawasam
Founder
Rhodes Scholar with an Oxford engineering PhD and University of Pennsylvania bioengineering background; previously CEO/founder of Simplifine (YC S24).
Founding Story
The founders started Cerulion after experiencing the cost and frustration of developing robots with slow, unreliable, difficult-to-debug tooling. Their initial vision was to make robot reliability something engineers could take for granted by providing fast communication, an easy cross-platform interface, better observability, dependable logging/replay, and less middleware boilerplate.
Business Model
Revenue Model
Open-source runtime and CLI under AGPL-3.0-only with a commercial license available; Hindsight and the hosted MCP server are closed-source early-access services accessed through a founder/demo booking.
Pricing Tiers
AGPL-3.0-only source-available/open-source distribution; the documentation states that a commercial license is also available.
Access is offered through a 30-minute booking; no public price is listed.
Target Markets
- Robotics startups and product teams
- Embodied-AI developers
- Industrial and research robotics teams
- Teams using ROS 2
- Robot developers needing real-time control, observability, replay, and CI verification
- AI-agent and automation developers
- Developing and deploying embodied-AI and autonomous-robot applications
- Real-time robot perception, control, and inter-process communication
- Investigating robot failures from recorded MCAP runs
- Replaying native robot recordings and verifying changes in CI
- Giving coding agents controlled tools to inspect and operate robot systems
- Running ROS 2 nodes on a faster, more observable transport
- Amazon Robotics engineers
- Boston Dynamics AI Institute engineers