How teams scaled GPUs: 3 real case studies
3 of the 73 engineering write-ups in our database mention GPUs in their stack. They come from 2 companies (Meta and Shopify) and were published between 2024 and 2026. Techniques that show up most often alongside GPUs: distributed training (2). It's often paired with PyTorch (2). 2 of the 3 state concrete scale numbers, for example "billions of product listings across millions of stores" (Shopify). Each card below links to our short summary and the original post.
Cross-store product clustering across millions of stores
Shopify groups the same real-world product sold by different merchants under one Universal Product Identifier. Embedding nearest-neighbor search finds plausible pairs across billions of listings, then three scorers of rising cost judge them: a CPU GBDT (L1) over trillions of pairs, a fine-tuned multilingual E5 cross-encoder (L2), and an image-aware reranker (L3) fed from prioritized GPU queues. A bounded union-find splits the graph into single-machine components, and HAC with a learned GBDT merge rule builds high-precision clusters daily, with an LLM judge validated against human labels supplying training data and cached scores reused between builds.
Meta's Generative Ads Model (GEM): Scaling the Central Model and Compute Engine for Frontier Ads AI
Meta replaced separate per-surface ad prediction systems with GEM, one central recommendation foundation model trained across all of its ad surfaces. A new LLM-style architecture (Kunlun) showed LLM-like scaling with 2.5x more model complexity, while custom low-precision kernels and 5D parallelism doubled end-to-end training efficiency to 20-25% MFU at 4x the training FLOPs across thousands of GPUs. Post-training transfer (LoopFM) and elastic pre-computation (Solaris) pass GEM's gains to real-time serving models without extra latency.
How Meta trains large language models at scale
Meta covers the hardware, networking and reliability work needed to train LLMs across very large GPU clusters.