Athena OS is an open-source platform for labeling audio and images for AI. It brings together speech transcription and speaker labeling with image annotation for object detection and segmentation. AI assistance can help draft or speed up repetitive labels; people review the results and decide what belongs in a dataset.
I’m Ts. Dominic Mathew, a Sarawakian tech founder currently based in Kuala Lumpur. I’m building Athena OS around a practical problem: AI projects need useful, well-organized data before they can produce useful results. That includes data from local languages and dialects, as well as images that need carefully drawn labels.
Why build a data labeling platform?
Every AI model depends on examples that describe what it should recognize. A speech dataset may need words linked to timestamps and speakers. An image dataset may need boxes around objects or masks that trace their shapes. Creating those labels by hand takes time, and inconsistent labels can make a dataset harder to use.
Athena OS is designed to help teams create and review these labels without making machine-learning code the first step. The goal is to make the labeling process clearer and more repeatable, while keeping people involved in decisions about the data.
Labeling Bahasa Sarawak audio
One Athena OS workflow is for audio labeled as Bahasa Sarawak. The interface brings the recording and transcript together, with timestamped segments and speaker labels. A transcriber can listen to a segment, move through the recording, edit the transcript, and review changes. The screen also includes an auto-transcription action.
Dialect speech data needs careful human review. Local expressions, pronunciation, spelling choices, and context can all affect a transcript. AI transcription can provide a starting point, but reviewers who understand the speech should check the words and timing before the labels are used in a dataset.
This workflow is useful for teams building speech datasets for a specific dialect, language community, or domain. It gives them a place to organize audio and its labels rather than treating a transcript as an unstructured block of text.
Annotating images for object detection and segmentation
Athena OS also includes an image annotation workspace. A labeler can choose a class and draw a bounding box around an object, or create a more detailed mask when the object’s outline matters.
- Object detection boxes mark where an object appears in an image.
- Segmentation masks mark which pixels belong to an object or region.
- Human review helps catch incomplete, misplaced, or inconsistent labels before export or training.
AI-assisted suggestions can reduce repeated work when a compatible model is available for the task. The labeler still needs to inspect and correct suggestions: an AI-generated label is a proposal, not automatically reliable ground truth. If no model is attached, the annotation workspace remains useful for manual labeling.
What does AI auto-labeling mean in Athena OS?
AI auto-labeling uses a model to propose labels for data, such as a transcript or an object region in an image. In Athena OS, these suggestions are intended to help with repetitive work. The person labeling the data remains responsible for reviewing and correcting them.
The exact workflow depends on the data type and whether a suitable model is connected. For example, an image annotation screen can support manual boxes and masks even when no image model is attached. Clear status and review steps help users understand whether they are working with model suggestions or creating labels themselves.
Why open source?
I want Athena OS to be useful to people who are building datasets in different fields and communities. An open-source approach gives teams a way to inspect the platform, adapt it to their workflow, and contribute improvements. It also makes data labeling a more accessible starting point for people who have domain knowledge but may not have a machine-learning engineering team.
My perspective comes from both places that matter to this work: I’m Sarawakian, and I’m currently based in Kuala Lumpur. I want to build practical AI tools that can serve local needs and connect with a wider developer and research community.
Frequently asked questions
What is Athena OS?
Athena OS is an open-source data labeling platform for audio and images. It supports speech transcription and speaker labeling, along with image annotation for object detection and segmentation.
Can Athena OS label a specific dialect?
Athena OS can organize audio transcription workflows for a specific dialect, including timestamped transcript segments and speaker labels. People familiar with the speech should review the transcription for accuracy.
Does Athena OS support object detection and image segmentation?
Yes. Labelers can use bounding boxes to mark objects for detection and masks to outline image regions for segmentation.
Does AI auto-label every image or audio file?
No. AI assistance depends on the workflow and whether a suitable model is connected. When the platform proposes labels, people should review and correct them. Manual annotation remains available when no model is attached.
Where is the founder of Athena OS based?
I’m Ts. Dominic Mathew — a Sarawakian tech founder currently based in Kuala Lumpur, Malaysia. Athena OS is intended for people and teams working with audio and image data wherever they are.
How does Athena OS relate to assisted CV labeling?
Athena OS is the open-source platform overview for audio and images. For the assisted computer-vision workflow—label one example, then review AI drafts across an image set—read Athena: AI-Assisted Data Labeling Without Code.
Building better datasets, one review at a time
Athena OS brings dialect audio labeling and image annotation into one open-source effort. The point is not to remove people from data work. It is to help them create useful labels with less repetitive effort, clearer review, and more control over what becomes training data.
Continue with the assisted labeling workflow, or explore Athena OS if you work with dialect speech, image datasets, or AI tools for local communities.