Loading AI Lab

Building the Future of AI from Vietnam

We are an artificial intelligence research company headquartered in Ho Chi Minh City, dedicated to pushing the boundaries of AI through years of rigorous research, thousands of experiments, and a strong commitment to responsible innovation.

Explore Contact Us
Abstract visualization representing AI research and computation
Let the data speak

Scale itself is proof of strength.

Our commitment is reflected not only in our lofty aspirations, but also in the depth and breadth of our work over the years.

50,000+
Logged experiment & evaluation records
2000+
Dataset organization and analysis
30+
Active research directions
340
Average number of experimental cycles
96%
Experiments reproducibility
6
Standard research cycle (months)
What we commit to

Four commitments that shape every cycle of our work

These are not pillars we describe once and forget — they come up in nearly every working session, and they decide where we slow down, where we push, and where we say no.

01 / Commitment

Patience over hype

We choose long cycles, repeated review, and steady iteration over chasing short-lived attention. Most of the value we deliver comes from quiet effort accumulated over months.

02 / Commitment

Local relevance

We stay close to the contexts we know. Vietnamese language, regional habits, and the constraints of small teams inform every decision about what is worth building and what is not.

03 / Commitment

Honest reporting

We describe progress as we see it — with the trade-offs, the limitations, and the things still unsolved. Calibrated language matters more to us than appearing certain.

04 / Commitment

Care for the human side

Behind every release there are real people doing the work and real people using the result. We design our cadence so neither group is asked to absorb avoidable harm.

Our work

Research spanning multiple cutting-edge fields

From natural language understanding to computer vision, our team works across the entire spectrum of AI to solve truly important problems.

Server room representing AI infrastructure

Infrastructure and Scale

Build the computing infrastructure needed for efficient training and deployment of AI systems. Manage over 500 GPU nodes and have independently developed an experiment management platform.

Cybersecurity and digital trust concept

AI Safety and Alignment

Dedicated to ensuring the reliable operation of AI systems and their alignment with human values. The security team has developed an assessment framework covering 200+ security scenarios.

A quiet workspace with notebooks and warm light, suggesting careful, patient work
Quiet work, real outcomes

Most of our progress happens away from the spotlight

The visible parts of our work — a release note, a short demo, a published article — are only the surface of much longer effort. Below that surface is a continuous practice of reading, sketching, debating, measuring, and rewriting that rarely gets shown.

We invest in that invisible layer on purpose. It is where most quality decisions are actually made, and it is what allows the visible parts to feel calm, considered, and dependable instead of hurried.

  • Long preparation cycles for any meaningful direction
  • Internal review sessions where disagreement is welcomed
  • Patient documentation of what worked and what did not
  • Deliberate pauses to reconsider direction when needed

A normal week here looks more like a library than a launchpad — and we think that is how it should be.

Our long promise
We will not pretend to be further than we are, we will not move faster than care allows, and we will keep showing up for the work even when no one is watching.
— The ZENTIX team