Computational efficiency
Understanding where computation, memory movement, and scheduling limit performance.
Open questions
- When does moving data cost more than computing on it?
- Which optimizations remain useful across different hardware?
Research areas
Our work begins with questions about how systems behave. We explore them through measurements, prototypes, and repeatable experiments.
Inference systems
Studying the movement, storage, and reuse of model state during inference.
Open questions
- When is moving KV cache cheaper than recomputing it?
- What changes when inference work spans multiple nodes?
Document processing
Extracting text and layout from real documents, including Thai and mixed-script material.
Open questions
- How well does synthetic data transfer to real Thai documents?
- How should text and layout errors be measured together?
Developer infrastructure
Designing small, reliable systems for access, metering, and product delivery.
Open questions
- How can usage be metered correctly across retries and failures?
- What infrastructure is necessary for a multi-product API platform?