https://sauravpanigrahi.com/tags/ml-systems/Recent content in Ml-Systems on Saurav PanigrahiHugoen-usFri, 01 May 2026 00:00:00 +0000https://sauravpanigrahi.com/reading/ml-systems/Fri, 01 May 2026 00:00:00 +0000https://sauravpanigrahi.com/reading/ml-systems/<p>Long-form references on training, infrastructure, and implementation practice.</p> <h2 id="training-systems">Training Systems</h2> <ul> <li> <p><a href="https://djdumpling.github.io/2026/01/31/frontier_training.html">Frontier Model Training Methodologies</a><br> Survey of open frontier training recipes and implementation choices.</p> </li> <li> <p><a href="https://jax-ml.github.io/scaling-book/">Scaling LLMs with JAX</a><br> Book-length treatment of distributed training practice.</p> </li> <li> <p><a href="https://arxiv.org/abs/2603.03276">Beyond Language Modeling: An Exploration of Multimodal Pretraining</a><br> From-scratch multimodal pretraining study with useful details on representation choices and scaling behavior.</p> </li> </ul> <h2 id="embeddings-and-retrieval">Embeddings And Retrieval</h2> <ul> <li><a href="https://blog.jxmo.io/p/how-to-train-the-best-embedding-model">How to Train the Best Embedding Model in the World</a><br> Detailed engineering writeup on embedding model training, label noise, verification, and dataset scale.</li> </ul> <h2 id="gpu-programming">GPU Programming</h2> <ul> <li> <p><a href="https://www.gpumode.com/home">GPU MODE</a> Community and resource hub for GPU programming.</p>https://sauravpanigrahi.com/reading/tool-use-and-agents/Fri, 01 May 2026 00:00:00 +0000https://sauravpanigrahi.com/reading/tool-use-and-agents/<p>Long-form references on tool use, agent environments, and reliability loops.</p> <h2 id="agent-environments">Agent Environments</h2> <ul> <li> <p><a href="https://openai.com/index/harness-engineering/">Harness Engineering</a><br> Useful framing around agents as systems shaped by environments, specs, feedback, and reliability loops.</p> </li> <li> <p><a href="https://blog.cloudflare.com/code-mode/">Code Mode</a><br> A concrete argument for exposing tools through code interfaces rather than forcing every step through chat-level tool calls.</p> </li> <li> <p><a href="https://mksg.lu/blog/context-mode">Context Mode</a><br> A useful pattern for keeping agent context manageable when tools produce large or noisy outputs.</p> </li> </ul> <h2 id="training-time-semantics">Training-Time Semantics</h2> <ul> <li><a href="https://arxiv.org/abs/2603.01209">Agents Learn Their Runtime</a><br> Study of persistent versus reset Python interpreters in CodeAct-style training.</li> </ul> <h2 id="engineering-practice">Engineering Practice</h2> <ul> <li><a href="https://kanyilmaz.me/2026/02/25/1000x-engineer.html">AI Gave Birth to the 100x Engineer</a><br> Long case study on compounding agent workflows with test harnesses and supporting tools.</li> </ul>