Synaptic Radio

Principal ML engineer focused on foundation models, agentic systems and scientific AI. I write about the math, systems and engineering judgement behind reliable ML workflows.

Home
Archives
Categories
Tags
Resume
Search
1. Dark Mode

Search

Categories

Distributed Training Evaluation and Experimentation Inference and Serving Post-Training and Alignment GPU Systems and Performance Engineering Software Architecture Tensor Parallelism Agents Deep-Learning Mechanisms Research Methodology

Tags

Calibration Evaluation Llm Confidence Distributed Systems Megatron-LM PyTorch Tensor Parallelism Uncertainty Actors

Categories

1 page

Scientific AI

A Better Benchmark Is Not a Better Scientist

Featured image of post A Better Benchmark Is Not a Better Scientist

© 2026 Anshuman Sahoo