Geodd

Geodd is an AI inference platform offering serverless endpoints, dedicated inference, and GPU clusters for production workloads. The API is OpenAI SDK-compatible, so switching providers requires changing one line of code. Geodd applies inference optimizations at the model and runtime layers (custom CUDA kernels, disaggregated prefill/decode, KV cache routing, FP8/FP4 quantization) to increase throughput without hardware upgrades. The platform claims 25-50% more throughput on existing GPU fleets and 2-3x faster generation via adaptive speculative decoding. Primary region is North America East (500+ GPUs), with EU and APAC regions coming.