Join Us Join Us Join Us

Author: Kayosan

May 1, 2026

When One Model Isn’t Enough: Building Multimodal Adult Content Detection for Mercari B2C

Abstract Every B2C product listing on Mercari must be screened for adult content before it reaches buyers. We built a multimodal ML pipeline that analyses both product images and listing text to make these decisions at scale. This post describes the system’s design: a custom PyTorch fusion model that combines MobileNet V2 image embeddings with Japanese BERT text embeddings, run in parallel with third-party API, with an OR gate combining their outputs. We cover our choice of fusion strategies (concatenation MLP vs. cross-attention), why we decided to run two independent classifiers rather than relying on one, and what we found when we attempted to improve performance by sub-classifying adult contents into finer-grained categories. We also outline the offline evaluation framework used to validate these decisions. Content Moderation on a B2C Marketplace Managing the scale of Mercari’s B2C catalog involves screening millions of listings, where every entry consists of seller-uploaded imagery and unstructured Japanese text (titles, descriptions, and category metadata). Ensuring these listings adhere to our safety policies is paramount, with adult content detection representing one of our most critical moderation challenges. The fundamental difficulty lies in the fact that “adult content” is a spectrum rather than a discrete binary class. On one end, we find unambiguous violations: explicit imagery or text that is clearly prohibited. At the other end, however, are edge cases where the content is benign in isolation but becomes problematic when combined. For instance, a piece of lingerie photographed on a mannequin is a standard product shot. The same garment, described with sexually suggestive language, shifts the listing into a different category. An art book containing classical nude paintings is a legitimate product. A cropped detail from one of those paintings, listed without context, raises different questions. This nuance implies that relying on a single modality, whether… <a class="more-link" href="https://about.in.mercari.com/news/blog/when-one-model-isnt-enough-building-multimodal-adult-content-detection-for-mercari-b2c/">Continue reading <span class="screen-reader-text">When One Model Isn’t Enough: Building Multimodal Adult Content Detection for Mercari B2C</span></a>

Published
Categorized as Blog

February 4, 2026

Optimists, Pessimists, and Realists: Leading in an AI-Native World – Sanu Satyadarshi

As Mercari continues to scale its Artificial Intelligence capabilities, we’ve had the opportunity to work closely with engineers, managers (both technical and non-technical), and leaders across the organization. One thing has become immediately clear: almost everyone has a strong opinion about how AI will change our world, or at least how it will change their work. These perspectives are often deeply polarized. Some see AI as the most transformative shift since the internet. Others view it with skepticism or concern. Both reactions are understandable. What matters most, however, is how organizations learn to navigate these differing worldviews while continuing to build reliable, production-grade systems. To make sense of these dynamics, one useful way to look at AI adoption inside organizations is through three broad cultural lenses: Optimists, Pessimists, and Realists. This framing isn’t meant to judge or rank perspectives. Instead, it helps leaders understand motivations, sources of friction, and how to channel disagreement productively. Part I: Understanding the Three Cultures The Optimists: The Force Multipliers Optimists are often found among senior leadership and engineers who are directly building AI- and LLM-based products. Their energy is infectious. Conversations with this group are often filled with statements like, “This is the biggest shift since the internet.” From their perspective, AI is a genuine force multiplier, work that once took weeks can now be done in days, sometimes hours. While this is an oversimplification, the underlying momentum is real. What defines Optimists is not uniform motivation, but consistent enthusiasm. They are excited by faster product cycles, new technical possibilities, and step-changes in how businesses operate. In many organizations, they become champions of change, pushing teams to think bigger and move faster. At the same time, optimism without intellectual humility can become a liability. Healthy ambition must be paired with rigorous questioning. The strongest… <a class="more-link" href="https://about.in.mercari.com/ai-native/optimists-pessimists-and-realists-leading-in-an-ai-native-world/">Continue reading <span class="screen-reader-text">Optimists, Pessimists, and Realists: Leading in an AI-Native World – Sanu Satyadarshi</span></a>