PinSieve: Production Selective VLM Serving and a Governed Memory Flywheel for Enterprise Content-Quality Triage

Enterprise AI agents in production often need to be bounded, stateful, observable, and governable rather than fully autonomous. We present PinSieve, a production case study in a large-scale content-quality pipeline. Its deployed component is a selective vision-language-model (VLM) Serving Agent that operates only on the grey-zone slice left unresolved by lightweight upstream models, exposes a scalar routing score online, and preserves controlled human escalation. On this slice, the deployed system filters 2.05x more non-actionable items than the previous production module while slightly reducing estimated miss rate; after promotion, it improves review productivity by 25.7%, reduces normalized operating cost by 16.2%, and moves signal delivery from next-day to same-day. We then study maintenance through a governed memory flywheel under selective feedback, where escalated items are reviewed by default and auto-passed items are labeled mainly through audit sampling. Feedback Memory records routing traces, observation paths, audit propensities, and replay metadata for evaluation and debugging. The Data Curation Agent uses a bounded proposal-verifier loop over representative, uncertainty, recency, and fresh-review replay, with positive-rate and score-bin guardrails before batch acceptance. In chained monthly refresh over six months of production data, this design reduces average FNR@50% from 17.73% under representative random replay to 13.29%. A Reasoning Review Agent audits teacher-generated rationales and supports keep/repair/drop decisions. Production claims are attributed only to the deployed Serving Agent; replay and rationale-review results are offline or sampled-governance evidence. The same serving-agent recipe has been adopted to several additional internal signals, suggesting transferability beyond one task.

Deep-learning Causal Retrieval Optimization for Efficient e-commerce Distribution in Pinterest

Pinterest is where people turn inspiration into action as users browse ideas, then take steps toward realization, often by discovering shoppable content. To support this journey, we must distribute commerce content when it helps, not when it distracts. We frame this as a causal decision of triggering shopping candidate generators in early retrieval and deploy a production system at Pinterest that learns personalized and contextualized triggering policies. A deep multi-task model jointly predicts outcomes and uplift of multiple events, trained with a doubly-robust pseudo-outcome alongside calibrated outcome losses for stable, single-robust uplift learning. A randomized data logging supplies counterfactual coverage, and the model is evaluated by both regular and reverse metrics for full assessment. A linear-time offline replay is designed to select thresholds and forecast policy impact with extremely high consistency with online results. For productionization, the model runs in parallel with remote retrieval calls without end-to-end latency regression. At web scale, we cut shopping triggers by up to 85% while holding key shopping sessions neutral, improving important total sessions (+0.26%) and Pin saves (+1.10%), with significant infrastructure savings. By unifying deep causal learning with reliable offline replay and demonstrating production-grade deployment, this work provides a generally practical recipe for early-retrieval optimizations in modern cascading recommenders beyond shopping, aligning exploration and cost with user intent at scale.

MatchLM2Lite: A Scalable MLLM-to-Lite Framework for Reproduced Content Identification

Content moderation is critical for online video platforms to ensure content safety, protect creators, and sustain positive user experiences. Beyond filtering harmful content, platforms must guarantee content authenticity at scale so that users are exposed to diverse, original videos rather than low-value reproductions. We present MatchLM2Lite, a real-time, production-grade reproduced content identification (RCI) system that leverages the powerful understanding of a multimodal large language model (MLLM) distilled into a small and fast-inference model. Our system jointly models video, audio, and text signals, operating on pairs of videos to produce fine-grained reproduction scores. The system comprises two modules, MatchLM and MatchLite, and a two-stage training recipe. First, our high-capacity MLLM, MatchLM, serves as a teacher model to define the upper bound of RCI performance. Its capabilities are then distilled into a compact student model, MatchLite. This design allows MatchLite to deliver low-latency, high-throughput inference on video pairs while preserving much of MatchLM's accuracy, making it suitable for integration into real-time recommendation systems. MatchLM achieves an F1-score improvement of +8.57 compared to our previous production model. After knowledge distillation, MatchLite retains a +6.55 gain in F1-score while reducing computational cost by 35x. Deployed at scale, MatchLM2Lite enables efficient, pairwise multimodal RCI, stably serving online traffic at high queries per second (QPS) with an end-to-end latency below 30 seconds. This system has reduced the reproduced video view rate on our platform by 2.5% without degrading user engagement, demonstrating its effectiveness in a large-scale production environment.

Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation

Collaborative filtering and graph-based recommendation models are highly effective because they leverage observed user interactions, but this dependence creates a fundamental cold-start challenge when newly added content has no interaction history. In Tubi's production retrieval system, this challenge is further constrained by the serving interface: new content must be assigned a standalone embedding immediately, and the model must also produce device embeddings suitable for approximate nearest-neighbor retrieval. We address this setting by formulating cold-start recommendation as an inductive graph-completion problem on a temporal bipartite device-content graph. We propose Shallow-RHS, an asymmetric link-prediction architecture in which the left-hand side (LHS) device tower leverages temporally valid watch-history message passing to capture collaborative signals, while the right-hand side (RHS) content tower is intentionally shallow with respect to the graph and encodes content solely from intrinsic features. The RHS tower does not use ID-based embeddings, content-side subgraphs, neighbor aggregation, or interaction-derived representations, forcing the content encoder to map intrinsic features into a collaborative-filtering-aware embedding space. After training, the learned content encoder generates embeddings for both warm and newly ingested content, enabling implicit graph completion through retrieval of warm surrogate neighbors. We further extend the same representation-completion principle to device cold-start by constructing cohort-based embeddings from demographic features. Large-scale online experiments demonstrate consistent relative improvements in content cold-start engagement, promotion speed, impression acquisition, and device cold-start engagement.