Prashant Anand

Machine Learning Engineer

Was an ML engineer at Mercari for 6 years building production ML systems for Japan's largest C2C marketplace. Led ML/NLP/LLM applications on Japanese text in Customer Support and Trust & Safety.

Led ML-driven product moderation system for Mercari Shops. Shipped automatic high-risk inquiry escalation, LLM-powered auto-replies, and intelligent inquiry routing.

B.Tech, IIT Delhi (2019). Speaker, PyCon JP 2024 and PyCon APAC 2023.

Work Experience

Machine Learning Engineer

Oct 2019 – Oct 2025

Mercari, Inc. • Tokyo, Japan

Progression: MG1 (10/2019) → MG2 (1/2021) → MG3 (7/2021) → MG4 (8/2024)

▸Led the design and implementation of the ML-driven product moderation system for B2C marketplace, Mercari Shops, defining the technical roadmap and coordinating with product and TnS teams.
▸Designed and deployed a real-time high-risk inquiry detection system using a stacked ensemble of fine-tuned ModernBERT and GPT-4.1, improving detection by 51% and cutting escalation time by 35%.
▸Shipped a DistilBERT-based auto-reply system that handles 13% of total inquiries, saving 2.200+ agent hours monthly. The system integrates 16 metadata features from 7 microservices.
▸Launched a GPT-3.5 auto-reply system for low-complexity inquiries, reducing manual replies by 39% and agent handling time by 6% within its target segment.
▸Optimized inquiry routing by productionizing fine-tuned Japanese DistilBERT model on 40+ skill categories, increasing assignment accuracy by 8% and reducing inter-agent transfers by 26%.