Thai Open AI Project

Building Open Artificial Intelligence for Thailand

งานวิจัยและพัฒนา Artificial Intelligence ตั้งแต่ Foundation Models, AI Agents, Computer Vision, Robotics, AI for Education ไปจนถึงระบบ AI ที่สามารถนำไปใช้งานจริง

Thailand Open Artificial Intelligence Research Initiative — Research. Experiment. Build. Share.

Dr. Kongkiat Hirankerd

Research led by

Dr. Kongkiat Hirankerd
ORCHESTRATORLLMVISIONSPEECHAGENTROBOTDATASETKNOWLEDGEREASONING

ACTIVE RESEARCH

12

AI MODELS

18

EXPERIMENTS

47

DATASETS

8

AI AGENTS

14

RESEARCH PAPERS

6

LIVE SYSTEMS

9

Featured Research

Research programmes in progress

Selected projects spanning protocol research, dataset construction, applied vision and education technology.

All projects →

Multi-Agent AI

UACP-HMA

Universal AI Communication Protocol for Heterogeneous Multi-Agent Systems — a transport-agnostic message contract that lets models from different vendors and runtimes cooperate on one task graph.

PythonFastAPIRedisPostgreSQL
RESEARCHUPDATED 2026-07-28

Multi-Agent AI

Thai Open AI Agent Lab

An observable environment where research agents run against real tasks with full logs of queue, latency, token usage and hand-off structure.

PythonReactTypeScriptSupabase
PROTOTYPEUPDATED 2026-08-04

Computer Vision

Drone Dataset TH

A dataset programme for detecting and classifying drones under Thai environmental conditions — tropical haze, monsoon light, dense urban clutter and long-range sky backgrounds.

PythonPyTorchYOLOSAM
EXPERIMENTUPDATED 2026-08-01

Computer Vision

Muay Thai Data & AI Platform

Fight analytics from video: fighter identification, pose estimation, strike detection, clinch analysis and multi-camera statistics for Muay Thai.

PythonPyTorchOpenCVMediaPipe
EXPERIMENTUPDATED 2026-07-19

AI for Safety

AI for EOR

Explosive Ordnance Reconnaissance decision support: visual identification, classification, hazard assessment and structured incident documentation for trained personnel.

PythonPyTorchPostgreSQLFastAPI
RESEARCHUPDATED 2026-07-30

Local & Sovereign AI

Local AI Research

Serving, routing and quantisation research for running Thai-capable models entirely on locally owned GPU infrastructure.

OllamaCUDADockerFastAPI
ACTIVEUPDATED 2026-08-08

AI for Education

Yenka AI Physics

AI-guided physics simulation and virtual laboratory with automatic experiment and question generation.

UnityBlenderPythonReact
BETAUPDATED 2026-06-22

AI for Education

Yenka AI Chemistry

Virtual chemistry laboratory with AI-generated experiments, safety-aware guidance and automated assessment.

UnityPythonReactTypeScript
PROTOTYPEUPDATED 2026-07-11

Research Areas

Research domains

Open research areas for experimenting with and developing open artificial intelligence technology, to build knowledge, technology and AI innovation that can be applied in Thailand.

Open research areas →

ระบบเอเจนต์หลายตัวทำงานร่วมกัน

Multi-Agent AI

Protocol-level research on how heterogeneous AI models negotiate, delegate and verify work across a shared communication layer.

AI Model A
Universal AI Communication Protocol
AI Model B
AI Model C

ปัญญาประดิษฐ์แบบท้องถิ่นและอธิปไตยข้อมูล

Local & Sovereign AI

Running capable models on Thai-owned infrastructure: local LLM serving, routing, GPU scheduling and offline-first deployment.

User
AI Gateway
AI Router
Local Models

คอมพิวเตอร์วิทัศน์และการรับรู้อัจฉริยะ

Computer Vision & Intelligent Sensing

Detection, tracking, pose and segmentation research for real deployment conditions — multi-camera, 360°, low light and long range.

Capture
Preprocess
Detection
Tracking

เอไอเพื่อความปลอดภัยและงานข่าวกรองวัตถุระเบิด

AI for Safety & Explosive Ordnance Intelligence

Decision-support research for reconnaissance teams: visual identification, classification, hazard assessment and knowledge retrieval. The system supports search, verification, classification, assessment and documentation only.

Field Capture
Visual Identification
Classification
Hazard Assessment

ปัญญาประดิษฐ์เพื่อการศึกษา

AI for Education

Adaptive tutoring, automatic experiment generation and virtual laboratories for physics and chemistry learning.

Learner Model
Adaptive Planner
AI Tutor
Virtual Laboratory

หุ่นยนต์และปัญญาประดิษฐ์เชิงกายภาพ

Robotics & Embodied AI

Perception-to-actuation loops, sensor fusion and autonomous field platforms operating with intermittent connectivity.

Sensors
Perception
World Model
Planner

Inside the AI Lab

AI Agent Network

Agents coordinate through an orchestrator, exchange evidence and verify each other's output before a result is returned.

Open network view →

Coordinator Agent

thai-open-router-1

WORKING

Assigning task to agents...

QUEUE
3
LATENCY
180 ms
TOKENS
12.4k
LAST
now

Research Agent

local-8b

WORKING

Searching research knowledge...

QUEUE
2
LATENCY
640 ms
TOKENS
48.1k
LAST
2s ago

Vision Agent

drone-detect-th

WORKING

Analyzing frame 2,841...

QUEUE
5
LATENCY
31 ms
TOKENS
LAST
now

Data Agent

pipeline-runtime

WORKING

Processing dataset shard 07...

QUEUE
1
LATENCY
1.2 s
TOKENS
LAST
4s ago

Reasoning Agent

cloud-reasoning

WAITING

Evaluating hypothesis...

QUEUE
0
LATENCY
2.4 s
TOKENS
91.7k
LAST
11s ago

Coding Agent

local-coder

ONLINE

Idle — awaiting task

QUEUE
0
LATENCY
410 ms
TOKENS
22.0k
LAST
1m ago

Simulation — agent activity shown here is demonstration data and is not connected to a live inference server.

Open Resources

Featured datasets & models

Versioned datasets and experimental models produced by the research programmes.

Drone Dataset TH

v0.4

EXPERIMENT

Drone detection and classification imagery captured in Thai environmental conditions with weather, camera and drone-type metadata.

IMAGES
collection ongoing
VIDEOS
multi-site
RECORDS
versioned releases
SIZE
LICENSE
Research use
UPDATED
2026-08-01
DetectionClassificationTrackingDrone typeEnvironmentWeatherCameraResolutionLabel confidence

Muay Thai Dataset

v0.2

EXPERIMENT

Annotated Muay Thai bout footage for fighter identification, pose, strike and clinch analysis.

IMAGES
frame-level
VIDEOS
multi-camera
RECORDS
round-indexed
SIZE
LICENSE
Research use
UPDATED
2026-07-19
Pose EstimationAction RecognitionTrackingFighter cornerStrike classClinch stateRound

drone-detect-th

EXPERIMENT

Detection model trained on Drone Dataset TH for small, distant aerial targets under tropical sky conditions.

BASE
YOLO family
PARAMS
~11M
TASK
Object Detection
LANGUAGE
QUANTIZATION
FP16 / INT8
HARDWARE
Single GPU or edge accelerator
LICENSE
Research use
VERSION
v0.4

Benchmark

  • Evaluationper-weather split
  • Target latencyreal-time edge

Limitations: Not validated for airspace-safety-critical decisions

thai-local-8b-q4

EXPERIMENT

Quantised local deployment configuration studied for Thai-language quality per watt on commodity hardware.

BASE
Open-weight 8B LLM
PARAMS
8B
TASK
Text generation
LANGUAGE
TH / EN
QUANTIZATION
Q4_K_M
HARDWARE
Consumer GPU, 8GB VRAM class
LICENSE
Depends on base model
VERSION
v0.2

Benchmark

  • Thai language scoreunder evaluation
  • Tokens/sechardware dependent

Limitations: Quantisation degrades long-form Thai reasoning; quantified in ongoing runs

Research Publications

Papers, reports and research notes

All publications →
TECHNICAL REPORT2026toap-tr-2026-01

UACP-HMA: A Universal Communication Protocol for Heterogeneous Multi-Agent AI Systems

Kongkiat Hirankerd

We describe a transport-agnostic message contract enabling models from different vendors and runtimes to cooperate on a shared task graph, with capability negotiation and provenance preservation across hand-offs.

Multi-Agent AIProtocolInteroperability
Related project
TECHNICAL REPORT2026toap-tr-2026-02

Drone Dataset TH: Data Collection Methodology for Tropical Aerial Detection

Kongkiat Hirankerd

A description of the capture protocol, AI-assisted labeling workflow and mandatory human verification stage used to build a versioned drone detection dataset for Thai conditions.

DatasetComputer VisionDrone Detection
Related project

Research Timeline

How the work developed

Full timeline →
  1. 2025-08 · Prototype

    Agent Lab prototype

    Instrumented multi-agent runtime with full execution traces.

  2. 2025-11 · Prototype

    Yenka AI Physics beta

    Virtual laboratory with simulation-grounded tutoring.

  3. 2026-03 · Experiment

    Muay Thai pipeline validation

    Pose-conditioned strike recognition experiments.

  4. 2026-06 · Research Started

    AI for EOR decision-support study

    Grounded retrieval architecture for reconnaissance support.

  5. 2026-08 · Deployment

    Thai Open AI Project platform v1

    Public research platform bringing projects, datasets and models together.

Research Collaboration

Work with the lab

We welcome collaboration with researchers, universities, government agencies, technology companies and open-source communities.

Research Updates

Latest from the laboratory

All updates →

2026-08-08 · Experiment Result

Quantisation sweep completed for Thai-language tasks

Full per-configuration throughput and VRAM figures are being prepared for the Local AI benchmark page.

2026-08-04 · Technology Update

Agent trace store now records full hand-off graphs

Every task in the Agent Lab now produces a replayable execution trace.

2026-08-01 · Dataset Release

Drone Dataset TH v0.4 versioning workflow live

Immutable releases with YOLO and COCO exports and complete capture metadata.

2026-07-19 · Research Progress

Clinch occlusion identified as dominant Muay Thai error source

Failure analysis is guiding the next round of pose-conditioning experiments.

Researcher

Dr. Kongkiat Hirankerd

Dr. Kongkiat Hirankerd

AI Researcher · Software Developer · Technology Researcher

Researcher profile
Artificial IntelligenceMulti-Agent AILocal AIComputer VisionAI for EducationAI SimulationRoboticsIntelligent SystemsDigital InfrastructureEmerging Technology