By Harshad Sane Ranker is one of the largest and most complex services at Netflix. Among many things, it powers the personalized rows you see on the Netflix homepage, and runs at an enormous scale. When we looked at CPU profiles for this service, one feature kept standing out: video serendipity scoring — the logic that answers a …
Read more “Optimizing Recommendation Systems with JDK’s Vector API”
Large language models (LLMs) perform well on general tasks but struggle with specialized work that requires understanding proprietary data, internal processes, and industry-specific terminology. Supervised fine-tuning (SFT) adapts LLMs to these organizational contexts. SFT can be implemented through two distinct methodologies: Parameter-Efficient Fine-Tuning (PEFT), which updates only a subset of model parameters, offering faster training …
The flexibility of Google Cloud allows enterprises to build secure and reliable architecture for their AI workloads. In this blog we will look at a reference architecture for private connectivity for retrieval-augmented generation (RAG)-capable generative AI applications. This architecture is for scenarios where communications of the overall system must use private IP addresses and must …
Read more “Designing private network connectivity for RAG-capable gen AI apps”
Authors: Harshad Sane, Andrew Halaney Imagine this — you click play on Netflix on a Friday night and behind the scenes hundreds of containers spring to action in a few seconds to answer your call. At Netflix, scaling containers efficiently is critical to delivering a seamless streaming experience to millions of members worldwide. To keep up with responsiveness …
Read more “Mount Mayhem at Netflix: Scaling Containers on Modern CPUs”
Large-scale commercial search systems optimize for relevance to drive successful sessions that help users find what they are looking for. To maximize relevance, we leverage two complementary objectives: behavioral relevance (results users tend to click or download) and textual relevance (a result’s semantic fit to the query). A persistent challenge is the scarcity of expert-provided …
Read more “Scaling Search Relevance: Augmenting App Store Ranking with LLM-Generated Judgments”
There’s a lot of excitement right now about AI enabling mainframe application modernization. Boards are paying attention. CIOs are getting asked for a plan. AI is a genuine accelerator for COBOL modernization but to get results, AI needs additional context that source code alone can’t provide.Here’s what we’ve learned working with 400+ enterprise customers: mainframe …
Read more “Learnings from COBOL modernization in the real world”
With the dawn of the gen AI era, businesses are facing unprecedented opportunities for transformative products, demanding a strategic shift in their technology infrastructure. A few years ago, PayPal, a digital-native company serving hundreds of millions of customers, faced a significant challenge. After 25 years of success in expanding services and capabilities, we’d created complexity …
Read more “PayPal’s historically large data migration is the foundation for its gen AI innovation”
Prior studies investigating the internal workings of LLMs have uncovered sparse subnetworks, often referred to as circuits, that are responsible for performing specific tasks. Additionally, it has been shown that model performance improvement through fine-tuning often results from the strengthening of existing circuits in the model. Taken together, these findings suggest the possibility of intervening …
Organizations and individuals running multiple custom AI models, especially recent Mixture of Experts (MoE) model families, can face the challenge of paying for idle GPU capacity when the individual models don’t receive enough traffic to saturate a dedicated compute endpoint. To solve this problem, we have partnered with the vLLM community and developed an efficient …
Something has shifted in the developer community over the past year. AI agents have moved from “interesting research concept” to “thing my team is actually building.” The prototypes are working. The demos are impressive. And now comes the harder question: How do we ship this? That question turns out to be a multi-part one. Agents …
Read more “A developer’s guide to production-ready AI agents”