Explore the decisions behind inference.
Understand search behavior, serving tradeoffs, and the model journey through interactive examples. No account, API key, or GPU required.
Search strategy playground
Replay actual TPE, Gaussian-process, and random sampler observations. Compare two objectives and two seeds.
The inference tradeoff
Change a latency ceiling, inspect Pareto points, and compare configurations within a workload.
A model’s journey
Follow evidence from discovery through pinned selection, execution, and optimization.
Recorded run replay
Replay imported inference runs, inspect throughput and p95 latency, and disclose missing provenance.
Real algorithms. Honest boundaries.
The lab includes offline simulations and recorded inference measurements. Simulations are labeled separately; measured runs retain their source context and disclose missing metadata. No demo establishes model compatibility or a performance advantage.
Read the methodology