lakesideai.dev

Run aerospace
trade studies by
describing them.

Real solvers. Real physics. Full audit trail.

Lakeside AI builds agentic infrastructure for aerospace engineering. Author concept studies (aero, structures, propulsion, mission) in plain language to an AI agent. The agent runs the actual analysis tools the way you would. Every result is traceable from requirement to figure.

Your tools, your physics, agent-driven, ready for your command.

01 / walkthrough

What it looks like.

A 90 second screencast: how a request becomes a finished concept study, agent-driven.

walkthrough.mp4
[ video · 16:9 placeholder ] 90 second walkthrough · coming soon

From request to result: an end-to-end concept study, agent-driven.

02 / pain

Let engineering do engineering.

Three ways the work that should take a day takes a week.

setup tax

You want to compare three wing concepts. That's a 30 minute decision and a two day setup: wrappers, OpenMDAO plumbing, fixing the design variables you typo'd, fighting the optimizer.

cross-tool tax

You want pyCycle to size the engine, OpenAeroStruct for the wing, OpenConcept for the mission. Three tools, three conventions, bespoke glue between them. The person who wrote that glue last time has moved on. You start over.

re-run tax

Last quarter you ran a study. Now you need to re-run it with one changed assumption. Where's the config? Whose laptop? What version of every tool?

We make this work disappear, without giving up the physics or the audit trail.

03 / how it works

Direct the study. Review the results.

Three steps from request to record.

01

Describe what you want.

Author a study in plain language or a YAML plan. Aero, structures, propulsion, mission. Single tool or composed.

02

The agent runs the real tools.

OpenAeroStruct, OpenConcept, pyCycle ready now. The same code you'd run locally, hosted and agent-ready. No approximations.

03

Every study writes its own record.

Every run captures what you asked for, what the tools did, what they returned, and the decisions in between. Replay any study. Compare two versions side by side. Pick up where someone else left off, weeks or months later, with the full picture intact.

provenance.mp4
[ video · 16:9 placeholder ] provenance walkthrough · coming soon

Provenance, by default. Every run, every decision, traceable.

04 / who

Who this is for today.

Two kinds of teams getting value from this now.

concept-design and advanced-design teams

Evaluate ten concepts in the time it used to take to evaluate two. Same physics, full defensibility, fraction of the engineer-hours.

research labs and small aerospace teams

Open-source solvers you already trust, with the orchestration and provenance layer you've been building yourself. Hosted, agent-ready, no glue code.

Not a solver replacement. Not a CAD tool. Your studies, now as a connected record you can search, replay, and build on.

05 / the technical product

Under the hood: The Hangar.

The open-source MCP server suite that makes this work.

The Hangar is our open-source MCP server suite. The technical foundation that makes this work. It exposes OpenAeroStruct, OpenConcept, and pyCycle as agent-callable tools, with a YAML-first OpenMDAO plan runner (omd) for composing studies across them. Every run produces a versioned envelope, a provenance graph, an N2 diagram, and a reproducible record.

hangar/oas Aerostructural analysis and optimization (OpenAeroStruct).
hangar/ocp Aircraft conceptual design and mission analysis (OpenConcept).
hangar/pyc Gas turbine engine cycle analysis (pyCycle).
06 / contact

Talk to us.

We're working with research labs, concept-design teams, and small aerospace organizations on early studies. If you have a problem you'd like to run through this, we'd like to hear about it.