InfraLens

A clear starting point for learning AI infrastructure.

Tools

Estimate before you choose.

Small calculators for building system intuition. They estimate orders of magnitude, not exact production capacity.

KV Cache Calculator

What it estimates: KV Cache memory for a batch and context length.

When to use it: before increasing context length, concurrency, or KV precision.

Attention Cost Estimator

What it estimates: naive attention score elements and memory.

When to use it: when sequence length or visual tokens make attention the bottleneck.

Video Token Count Estimator

What it estimates: image tokens and spacetime tokens.

When to use it: before changing resolution, frame count, patch size, or temporal patching.

Quantization Memory Estimator

What it estimates: parameter memory before and after quantization.

When to use it: when comparing FP16/BF16 weights with int8, int4, or scale overhead.