» Tag
machine learning
9 postsWhy price per 1M tokens is a misleading AI metric
Comparing AI models by price per 1M tokens can mislead teams. Tokenizer differences and chain-of-thought efficiency matter far more than the sticker price per token.
aillmmachine learningbenchmarkingFrom a 48-hour PDF rabbit hole to a repeatable research pipeline
An engineer's 48-hour struggle extracting data from academic PDFs led to an automated, auditable research pipeline with measurable gains in speed and reliability.
pdf-processinglayoutlmmachine-learningresearch-automationSentry's New AI Model Cuts Issue Grouping Errors in Half
Sentry upgraded its AI-powered issue grouping model, cutting duplicate issue creation by 20% and halving the rate of incorrect merges in production.
sentrymachine-learningobservabilityerror-trackingDon't Let the LLM Speak, Just Probe Its Hidden State
LLMs already encode the answer in hidden states before generating text. A tiny probe turns this into fast, cheap, calibrated zero-shot classifiers without generation.
llmnlpmachine-learningclassificationNetflix Unveils Slow-Fast Architecture for Notifications
Netflix built a hierarchical notification system inspired by Kahneman's fast/slow thinking model, decoupling long-term pacing strategy from real-time message selection.
netflixpersonalizationrecommendation-systemsmachine-learningHow to Design a Cancer Vaccine: Neoantigens and MHC Science
UNC's PIRL lab founders discuss neoantigen detection, MHC prediction, and the engineering challenges of personalized cancer immunotherapy in a three-hour deep dive.
cancer vaccinesimmunotherapymachine learningbioinformaticsNew study finds language models memorize about 3.6 bits per parameter
Researchers unveil a method to measure LLM memorization capacity, finding GPT-style models store roughly 3.6 bits of information per parameter, with implications for grokking and scaling.
llmmemorizationmachine-learningscaling-lawsBuilding a Multimodal Vector Search Platform for Product Catalogs
How a recommendation team built a multimodal vector search platform with Qdrant, tuned HNSW settings, and optimized GPU usage for embedding millions of products.
vector-searchembeddingsqdrantmachine-learningAI Inference Engineering: Inside the Prefill-Decode Split
A technical look at how LLM inference splits into compute-bound prefill and memory-bound decode phases, and the optimization techniques engineers use to scale them.
aiinferencellmgpu