Experiments & Lab
A log of prototypes, failed hypotheses, and technical explorations in machine learning and distributed systems.
Lab Experiment•Oct 14, 2023
Quantized LLaMA Routing Latency
Evaluating the latency and accuracy overhead of using a lightweight embedding model to route prompts to specialized local LLMs based on semantic intent.
Lab Experiment•Sep 22, 2023
Embedding Drifts in Long-Running Agents
Analyzing semantic degradation over time in long-running autonomous agents when adapting to shifting contextual windows.