A mixture-of-experts layer is 256 small matmuls. Why is a loop over cuBLAS calls slow, and what does a grouped GEMM do instead?

Routing hands each expert a different number of tokens, so the layer is a list of ragged matmuls rather than one. Three separate costs a loop pays, the measured share of compute that skewed routing wastes on tile padding, and what a single grouped launch fixes and what it cannot.

Updated Sep 2026 · Grounded in real AI infrastructure interview loops and written to a senior-engineer editorial bar, with every number worked and every diagram hand-built.

20 answers per topic instead of 10, plus saved progress and bookmarks · no cardor unlock all 283 remaining answers · ₹2,000 / $25

The concepts behind this question

Ranked by how closely each one overlaps this question's topic, so the first card is the thing to read if the answer above moved too fast.

Advanced

🕸️ Distributed Training🔒 Premium

Expert Parallelism for MoEA mixture-of-experts layer runs only a few of its experts per token, so the experts can be spread across GPUs and each token shipped to the ranks that hold its chosen experts. That shipping is an all-to-all in each direction, twice per layer per pass, and its cost plus the load imbalance between experts is what expert parallelism is really about.

Foundational

🧮 Open Weights & Serving Engines

Capacity Planning for Open-Weights FleetsPlanning a fleet for a sparse open-weights model works differently from planning one for a dense model, because memory follows total parameters and throughput follows active parameters, and those now differ by more than twenty times. The sizing goes in one direction only: from a traffic forecast to tokens per second, to replicas at a measured operating point, to GPUs, to racks and kilowatts. Doing it in the other direction, from an available GPU count, produces a fleet that fits the hardware rather than the demand.

Foundational

🧮 Open Weights & Serving Engines

Reading config.json to Size a Model You Have Never RunEvery Hugging Face model ships a config.json, and it contains enough to compute the weight footprint, the KV cache per token, the parallel degrees that divide cleanly and the minimum GPU count, before downloading a byte. Doing that derivation is a standard whiteboard exercise in serving interviews because it is exactly what an engineer does on the morning a new model lands, and the fields that matter are the same across every recent architecture.

Foundational

🧮 Open Weights & Serving Engines

Expert Parallel and All-to-All BackendsA mixture-of-experts model can be split two ways and the choice changes everything. Tensor parallelism shards each expert across GPUs, which keeps every GPU busy and reads every expert's shard on every token. Expert parallelism gives whole experts to whole GPUs, which reads only the selected experts but requires an all-to-all to route tokens to them and back. The all-to-all is the cost, its backend is a configuration choice matched to the interconnect, and expert load imbalance is what actually limits the result.

Explore the full curriculum →

UP NEXT ON YOUR JOURNEY

Next in this trackSelect the top k logits from a 128,000-token vocabulary on the GPU. What shape does the kernel take and why not just sort?Next in this trackYou added three local variables to a working kernel and it got 30 percent slower. Explain what happened and how you would confirm it.Next in this trackA training step runs at 20 percent model FLOPs utilization. Profile it and find where the missing time goes.

FEDITOR'S NOTE

Scored on separating launch overhead from underfilled tiles from load imbalance, on the tile-quantization arithmetic, and on knowing that grouped GEMM fixes scheduling and not routing balance.

DISCUSSION · 0

No comments yet — be the first to share your approach.