General-purpose embeddings trained on profile similarity often misrank candidates in professional networks, where relevance is inherently complementary rather than symmetric. For example, job seekers usually engage with recruiters rather than peers with similar credentials. Current user recommendation systems are often built on profile similarity or social signals such as shared connections, which are effective in friend-oriented social networks where users tend to connect with similar people or within existing social circles. However, this does not hold in professional platforms like LinkedIn, where interactions are often goal-driven and cross-role interactions, requiring matching based on complementary intents rather than similarity or social proximity. In this paper, we propose Preference-Aligned Intent Retrieval, a retrieval framework that is designed to model user intent preference instead of profile similarity. Our approach is inspired by the preference-learning perspective of Direct Preference Optimization, but instantiates preference learning as an exposure-conditioned multi-negative contrastive objective over embeddings. This formulation simplifies training and scales efficiently to large candidate sets without requiring an explicit reference model. Furthermore, we incorporate user interaction history by aggregating click histories with positional encoding to capture temporal order, which enables constant-cost updates of user representations over a fixed history window, allowing new interactions to be incorporated without re-encoding history. As a result, personalization can be updated efficiently using lightweight CPU operations without repeated LLM inference during serving. We evaluate PAIR on LinkedIn's Premium retrieval task. A 0.6B parameter model achieves a MAR@100 of 0.45.
Figures & tables
Figure 1. Preference-Aligned Intent Retrieval architecture. The model adopts a triple-tower design. The profile tower and document tower share parameters to encode query and candidate profiles. The history tower aggregates recent interaction embeddings using positional attention to capture short-term behavioral intent. The final query representation combines static profile identity and dynamic intent through normalized additive fusion, enabling efficient history updates without re-encoding the full profile through the LLM.
Model
ACC ↑
MAR@25 ↑
MAR@100 ↑
Baselines
Holistic-171
0.794
0.169
0.288
Off-the-shelf 0.6B embedding
0.771
0.100
0.192
PAIR fine-tuned from the 0.6B encoder
Two-tower
0.960
0.255
0.436
+ Max history
0.963
0.252
0.430
Table 1. Main results on Premium user retrieval. Off-the-shelf models are evaluated without fine-tuning. PAIR models are fine-tuned from the 0.6B embedding encoder. ACC is computed over 4 candidates consisting of 1 positive and 3 negatives. MAR@ k is evaluated on a fixed pool of 100 candidates.
Method
MAR@100 ↑
Update Cost
Operations
No history
0.436
–
Static profile only
Max pooling
0.430
O(Kd)
Element-wise max
Mean pooling
0.430
O(Kd)
Element-wise mean
GeM pooling
0.447
O(Kd)
Power mean
Attention
0.431
O(Kd)
Dot product + softmax + weighted sum
P-ATT
0.450
O(Kd)
Positional encoding + dot product + softmax + weighted sum
Table 2. History aggregation tradeoff. All history methods aggregate K=10 precomputed history embeddings with 1024 dimensions. None require LLM re-inference during history updates.
Dim
MAR@100 ↑
Rel. Storage
Rel. Latency
Intended Surface
200
0.421
0.20 ×
0.35 ×
Candidate pre-filtering
256
0.432
0.25 ×
0.40 ×
Latency-sensitive channels
512
0.444
0.50 ×
0.65 ×
Standard retrieval
1024
0.450
1.00 ×
1.00 ×
Premium Browsemaps
Table 3. Matryoshka dimension-quality tradeoff. All variants are produced by a single PAIR model trained with joint Matryoshka loss. Relative storage is normalized to the 1024-dimensional embedding. Relative latency is based on dimension-dependent retrieval cost and fixed serving overhead.
Metric
Relative Change
Click-through rate (CTR)
+7.5% ∗
Premium retention rate
+2.1% ∗
Page load latency
+0.1%
Recommendation coverage
-0.2%
Table 4. Online A/B test results on Premium Browsemaps. Treatment: PAIR with P-ATT history augmentation at 1024 dimensions. Control: production Holistic-171 embedding. Both arms use the same downstream ranker. Metrics are reported as relative changes over the control group. Statistical significance under the production experimentation framework is indicated by ∗ .
In large-scale recommendation systems like the LinkedIn Feed, content generated by a member's network (connections and follows) makes up over 70% of impressions and engagement. It is therefore essential that the pre-ranking layer forwards the best possible few hundred candidates to the ranking layer. LinkedIn's professional knowledge graph carries engagement signals across both the first degree network (connections and follows) and the second-degree network: posts that a 1st-degree connection reacted to, commented on or reshared but did not author (a.k.a. stranger viral). Due to this fan out, the resulting candidate index exceeds one billion; selection of activities from the viewer's network narrows it down to roughly tens of thousands of activities that must be scored within a 120 ms p99 latency budget. We present Connected Content Retriever (CC Retriever), a pre-ranking system that scores these candidates with a full deep ranking model on GPUs at low latency. At its core is a sorted-search GPU primitive that joins dense graph affinity features (viewer to author) with document level features stored on the GPU at runtime in 5-10 ms. The shift to GPU served scoring enabled a 50x scale up of the ranking model's parameters and delivered a +2.5% lift in content time spent on the LinkedIn Feed in online experiments, significantly higher than the typical gains observed in LinkedIn Feed experiments. In this work, we describe the feature set we leverage from LinkedIn's economic graph and the model architecture used for scoring, with a particular emphasis on the online system that scales the stack.
Semantic Search on LinkedIn must retrieve relevant profiles from a corpus of hundreds of millions in response to natural-language queries such as "a fintech founder in Berlin who worked in payments." The deployed relevance policy is bottleneck-oriented: every active non-negotiable facet must be satisfied, and a pre-existing LLM Graded Relevance (GR) judge operationalizes this through a fixed min/median aggregation over facet grades. Cosine similarity instead averages evidence, letting a strong match on one facet mask failure on another, capping the recall of the first-stage (L0) retriever. We present a policy-aligned retrieval framework: embeddings are partitioned into eight category-supervised segments whose scores follow the same min/median rule at serving time; for multi-vector retrieval, this segment score is computed independently per tagged document slot and maximized across slots. A lightweight single-slot Stage-1 scorer generates high-recall candidates, while scale-invariant relative-norm gating keeps category activation consistent across training, evaluation, and serving. On 21K held-out queries, this representation improves offline relevance over a matched-capacity baseline, with gains broadly distributed across facet combinations. We serve this framework with a two-stage GPU architecture: an FP8 coarse ranker scores the full corpus, increasing per-shard capacity by 71% and Stage-1 matmul throughput by 36%, then an FP16 stage exactly re-ranks an oversampled candidate set, recovering 99.6-99.8% of full-FP16 recall at over 500 QPS per shard replica. In a member-randomized A/B test, exploratory-query Precision@10 under the unchanged GR judge rises from 63.7% to 79.0% and navigational Precision@1 from 65.5% to 74.7%, with a blinded human evaluation independently confirming the Precision@10 gain.
Modern recruitment platforms operate under severe information imbalance: job seekers must search over massive, rapidly changing collections of postings, while employers are overwhelmed by high-volume, low-relevance applicant pools. Existing recruitment recommender systems typically rely on keyword matching or single-stage semantic retrieval, which struggle to capture fine-grained alignment between candidate experience and job requirements under real-world scale and cost constraints. We present Synapse, a multi-stage semantic recruitment system that separates high-recall candidate generation from high-precision semantic reranking, combining efficient dense retrieval using FAISS with an ensemble of contrastive learning and Large Language Model (LLM) reasoning. To improve transparency, Synapse incorporates a retrieval-augmented explanation layer that grounds recommendations in explicit evidence. Beyond retrieval, we introduce a novel evolutionary resume optimization framework that treats resume refinement as a black-box optimization problem. Using Differential Evolution with LLM-guided mutation operators, the system iteratively modifies candidate representations to improve alignment with screening objectives, without any labeled data. Evaluation shows that the proposed ensemble improves nDCG@10 by 22% over embedding-only retrieval baselines, while the evolutionary optimization loop consistently yields monotonic improvements in recommender scores, exceeding 60% relative gain across evaluated profiles. We plan to release code and data upon publication.