Built in Thailand, for the teams a foreign cloud cannot serve.
Axentx is a deep-tech company in Bangkok. We build an engineering agent and a model for teams that run regulated or air-gapped systems, where the code an agent reads can never reach a network the customer does not control.
The most capable agents cannot go where this work lives.
A coding and monitoring agent works best with full context: your source, your secrets, your telemetry. For a bank, a hospital, or a government team, sending that to a foreign cloud is not a hard decision. It is simply not allowed.
So the frontier agents are off the table for exactly the teams that most need one. A model small enough to self-host, specialized on their domain, closes that gap. Open weights and cheaper inference make it practical now, on a single GPU.
Where things stand.
Founder-led
Started by an engineer who spent years running DevSecOps and SRE inside banking, and who built the training pipeline, the agent, and the dataset the model runs on.
Bangkok, Thailand
Close to the first market: Thai and Southeast Asian institutions bound by data-residency rules, who need an agent that stays in-country.
NVIDIA Inception member
We train on NVIDIA GPUs and deploy inference on NVIDIA hardware. Every sovereign customer runs the model on their own NVIDIA GPU.
Early, and building in the open path
Pre-seed, working with a small set of design partners. The goal is to publish the harness and the model, not to sell weights.
If this is your problem, we should talk.
Design partners, investors, or engineers who want to see it run. All of it reaches the same inbox.