lwatiAI
Sign in Start building
LOCAL-FIRST COMPUTATIONAL WORKSPACE

Build.
Compute.
Learn.

Build software, analyse data, conduct scientific work and learn through real computation — in one browser-based workspace designed to keep useful intelligence close to you.

Start building Explore Science
Local AIPythonDuckDB SQLRExportable work
Studio Local-first
Analyse crop yield by district, explain the result, then turn it into a dashboard.
01Datasetrainfall.csv
02DuckDB SQLPASS
03Python analysisPASS
04DashboardReady
The AI can plan. The runtime performs the calculation.
BUILDING WITH A WIDER ECOSYSTEM Member of NVIDIA Inception

Lwati AI has been accepted into NVIDIA's startup program for companies building AI products.

ONE PLATFORM / THREE WORKSPACES

Different kinds of work.
One computational core.

Move from an idea to code, from a question to evidence, or from a concept to an executable assessment without restarting your workflow in another product.

01
↗

Build

Turn ideas into working software. Edit real source, inspect changes, preview the result and export what you make.

AI-assisted development · source ownership
 02
↗

Science

Run Python, analytical SQL and R in the browser. Connect evidence, datasets, analyses, results and claims.

Pyodide · DuckDB-Wasm · webR
 03
↗

Learn

Learn by doing. Create assessments that can execute code and calculations instead of only checking selected answers.

JavaScript · Python · SQL · R
SCIENCE / DETERMINISTIC COMPUTE

The AI explains.
The runtime calculates.

Scientific answers should come from executed computation, not from a model guessing the arithmetic. Lwati routes work to real browser runtimes and preserves how the result was produced.

See Lwati Science ↗
PY
PythonCPython via Pyodide
LOCAL
SQL
DuckDBCSV · JSON · Parquet
LOCAL
R
webRStatistics and research
LOCAL
LOCAL-FIRST BY DESIGN

Your computer can be part of the cloud.

Not every useful action should require a remote inference call. Lwati uses local browser capabilities where they fit and keeps connected services explicit.

01

Local intelligence

Supported coding models can run with WebLLM and WebGPU. Capability checks matter: a working editor never depends on pretending every GPU is the same.

02

Local computation

Python, SQL and R can perform useful scientific work even when a generative model is unavailable.

03

Portable outputs

Source code, datasets, evidence, notebooks and research capsules are meant to move with the person who created them.

04

Connected when useful

Account saves, external evidence sources, deployments and heavier compute can use the network when the task genuinely needs it.

KNOWLEDGE SHOULD MOVE, NOT RESTART

Learn it.
Test it.
Build with it.

LEARN

Understand a concept.

Practice it through executable assessment.

↓
SCIENCE

Test it with data.

Run real computation and preserve the evidence.

↓
BUILD

Turn it into software.

Make the result useful to somebody else.

ZERO-INFERENCE PRODUCTION

Build with AI.
Run without it.

AI can help create an application without becoming a permanent tax on every form submission, report or page view. Ordinary workflows can remain ordinary software.

Why Lwati is built this way ↗
01AI during development
→
02Portable source
→
03Normal deployment
→
04Users
START WITH SOMETHING REAL

An idea. A dataset. A question. A skill.

Build a clinic management system.Analyse rainfall and crop-yield data.Teach Python using executable assessments.Investigate a public-health dataset.Build an offline field-data application.Run statistics in R without installing R.Turn research results into a public dashboard.Create a coding assessment that runs the answer.
OPEN TECHNOLOGIES / ORDINARY INFRASTRUCTURE
WebLLMWebGPUWebAssemblyPyodideDuckDB-WasmwebRLaravelReact

Use powerful browser standards where they fit. Escalate to stronger local or remote compute when the workload demands it.

QUESTIONS

What local-first actually means.

Does Lwati require a powerful GPU?+

No for the workspace itself. Local generative AI benefits from compatible WebGPU hardware, while deterministic scientific runtimes such as Python, SQL and R can operate independently. Stronger execution options can be used for workloads that exceed the device.

Does my code have to be uploaded?+

No. Browser-local projects can remain on the device. Account synchronization and connected services are separate choices and should be explicit.

Does software built with Lwati need Lwati to run?+

Not by default. Lwati is designed to create ordinary, portable source. If you deliberately add an AI-dependent feature, that feature will have its own runtime requirements.

Can Science work without the local AI model?+

Yes. Python, DuckDB SQL and R are deterministic runtimes. AI can assist with planning and interpretation, but the calculation itself does not need a language model.

What does local-first mean for privacy?+

It means the architecture starts from the assumption that useful work can remain on the user's device. External APIs, account sync and remote services still require network access and have their own data boundaries.

THE NEXT THING STARTS WITH YOU

What will you make,
discover or learn?

Start with an idea, a dataset, a question or a blank file.

Start building Open the Science story →