01Active
RAG experiments
Retrieval quality, chunking, hybrid search and grounded answer behavior.
AI engineer’s lab
A working shelf for experiments that are useful enough to share but not yet case studies.
02Active
LLM evaluation
Small evaluation sets and failure analysis for application-specific behavior.
03Ongoing
Model comparisons
Capability, latency and cost comparisons tied to defined product tasks.
04Exploring
Voice AI
Interaction patterns and system boundaries for useful real-time voice experiences.
05Research
Sinhala NLP
Fine-tuning, correction and evaluation for underrepresented language workflows.
06Ongoing
Prompt systems
Versioned prompts, structured output and regression-aware iteration.