Haoling Pu — ML systems engineer. CMU AI MS '27, Google intern, CUDA kernels, LLM agents, simultaneous speech translation.https://www.haolingpu.com/https://www.haolingpu.com/projects/ml-notes/https://www.haolingpu.com/projects/ml-notes/Everything between a prompt and a token, one mechanism at a time: an interactive map of the LLM serving stack, from attention and the KV cache to agents and GPU kernels.Fri, 02 Oct 2026 00:00:00 GMThttps://www.haolingpu.com/projects/sglang-omni/https://www.haolingpu.com/projects/sglang-omni/Contributor to the open-source serving framework for speech models. My work is on making the audio arrive on time, not just arrive.Fri, 02 Oct 2026 00:00:00 GMThttps://www.haolingpu.com/projects/agent-anchor/https://www.haolingpu.com/projects/agent-anchor/Keeping production AI agents from quietly getting worse: a loop that notices when quality drops, finds the cause, and checks the fix.Sun, 20 Sep 2026 00:00:00 GMThttps://www.haolingpu.com/projects/team-knowledge-agent/https://www.haolingpu.com/projects/team-knowledge-agent/The agent that knows what the team knows. Shipped for real at Google, for teams beyond my own.Sat, 15 Aug 2026 00:00:00 GMThttps://www.haolingpu.com/projects/llm-wiki/https://www.haolingpu.com/projects/llm-wiki/A wiki that writes itself: an agent reads sources, links pages, and cleans up its own contradictions. Karpathy's LLM Wiki idea, made real.Mon, 01 Jun 2026 00:00:00 GMThttps://www.haolingpu.com/projects/sparse-attention-blackwell/https://www.haolingpu.com/projects/sparse-attention-blackwell/A CUDA kernel that makes DeepSeek-style sparse attention fly on Blackwell. Passed every workload in NVIDIA's competition, dozens of times faster than the reference.Wed, 15 Apr 2026 00:00:00 GMThttps://www.haolingpu.com/projects/hybrid-rag/https://www.haolingpu.com/projects/hybrid-rag/Keyword search and embeddings, fused so each covers the other's blind spots, with a local open-source model doing the talking.Sun, 01 Mar 2026 00:00:00 GMThttps://www.haolingpu.com/projects/petfinder-pro/https://www.haolingpu.com/projects/petfinder-pro/Lost a pet? Post it, and let matching do the searching.Wed, 28 Jan 2026 00:00:00 GMThttps://www.haolingpu.com/projects/echoframe/https://www.haolingpu.com/projects/echoframe/Hand it a faded old photo, get it back in color half a minute later. Built for my family's albums.Mon, 01 Dec 2025 00:00:00 GMThttps://www.haolingpu.com/projects/simultaneous-mt-taf/https://www.haolingpu.com/projects/simultaneous-mt-taf/An early bet that translating in real time gets easier if you guess the future first.Thu, 23 Oct 2025 00:00:00 GMT