At TENS LAB, I’m building deterministic context representation and token compilation systems for language model inference. CONTEX uses a canonical intermediate representation to support KV prefix-cache reuse and deterministic inference across LLM runtimes.
I’m also developing CONTEXDB, an inference-native context storage engine designed around prefix-cache locality. My research benchmarks latency, compute, and token cost from prompt syntax overhead in frontier LLM architectures.
I’m still early in my learning, and I’m intentional about understanding how technology and systems work. I explore ideas through self-learning, research, and small experiments, with interests in system design and how technology scales in real environments.
