AI Research
& Engineering.
The Lab is where we test models, build prototypes and explore what AI can do next — from language and voice to vision and private infrastructure.
Large Language Models
Prompting, fine-tuning and evaluation of open and frontier models for real product work.
- Evaluation harnesses
- Domain adaptation
- Vietnamese language quality
AI Agents
Multi-step agents that plan, call tools and recover from errors — connected through MCP.
- MCP servers
- Tool orchestration
- Guardrails
Voice AI
Speech recognition, pronunciation scoring and natural voice interfaces.
- Speech-to-text
- Pronunciation feedback
- Text-to-speech
Computer Vision
Understanding images and documents: detection, OCR and visual search.
- OCR & documents
- Object detection
- Visual search
AI Automation
Turning repetitive operations into reliable, observable automated pipelines.
- Back-office workflows
- Data extraction
- Human review loops
Local AI / Private AI
Running models on private infrastructure so sensitive data never leaves your environment.
- Self-hosted models
- On-prem deployment
- Data privacy
AI Infrastructure
Serving, scaling and monitoring models efficiently — latency, cost and reliability.
- Model serving
- GPU scheduling
- Observability
Your models.
Your infrastructure.
Your data.
For organizations that cannot send data to third parties, we deploy and tune open models on private servers — with the same evaluation, monitoring and security as any production system.
Assess
Which tasks, which data, what privacy constraints.
Select
Benchmark open models on your own evaluation set.
Deploy
Serve on private GPU or CPU infrastructure, on-prem or in your cloud.
Operate
Monitoring, updates and continuous evaluation.
How we research.
Measure first
Every experiment starts with an evaluation set, so results are numbers, not impressions.
Private by default
We prefer architectures where sensitive data can stay on infrastructure you control.
Ship what works
Experiments that prove themselves graduate into our products and client systems.
Share what we learn
We publish notes, findings and open tools through Insights.
Want to run AI
on your own terms?
Talk to the Lab about private deployments, custom models or a research collaboration.