In large-scale recommendation systems like LinkedIn’s, the retrieval stage is critical for narrowing billions of potential candidates to a manageable subset for ranking. LinkedIn's feed now serves suggested content based on the topical interests of members, where 2000 candidates are retrieved from several million candidates with a latency budget of a few milliseconds and inbound QPS of several thousand per second. This paper presents a novel retrieval approach that fine tunes a large causal language model (Meta’s LLaMA 3) as a dual encoder to generate high quality embeddings for both users (members) and content (items), using only textual input. We describe the end to end pipeline, including prompt design for embedding generation, techniques for fine tuning at LinkedIn scale, and infrastructure for low latency, cost effective online serving. We share our findings on how quantizing numerical features in the prompt enables the information getting encoded in the embedding facilitating greater alignment between the retrieval and ranking layer. The system was evaluated using offline metrics and an online A/B test, which showed substantial improvements in member engagement. We observed significant gains among newer members, who often lack strong network connections, indicating that high-quality suggested content aids retention. This work demonstrates how generative language models can be effectively adapted for real time, high throughput retrieval in industrial applications.

With the recent approval of IEEE 802.11n standard and progress on IEEE 802.11s mesh standardization efforts, performance evaluation of such standards has gained acceleration. In particular, several wireless testbeds have been created to form wireless mesh networks (WMNs) in order to test various aspects of these new standards. However, none of these considered multi-hop performance of WMNs when 802.11n based nodes are employed. While IEEE 802.11s is geared for implementing multi-hopping capability among the nodes, current studies still assume IEEE 802.11a/g based nodes. This paper presents a performance evaluation of WMNs using both 802.11s (in conjunction with 802.11a/g) and 802.11n on a real academic testbed. While a draft version of 802.11s has been used for creating a linear WMN using 802.11a/g based routers, the same linear WMN using 802.11n has been created using virtualized interfaces at the IP layer. Using such WMNs, throughput performance of TCP and UDP at both the 2.4 GHz and 5 GHz spectrums with different number of hops has been examined. The experiment results have shown that multi-hopping significantly degrades the expected performance of IEEE 802.11n.