Hybrid Research Engineer – Applied AI Research

at Mercari, Inc.

• Job title: Research Engineer – Applied AI Research
• Job type: Full Time
• Experience level: Mid level, Senior
• Salary: Not specified
• Degree requirement: Postgraduate Degree
• Tech skills: Python, PyTorch
• Location requirements: Hybrid, Minato City, Japan

About the role

• Research Engineer position at Mercari focused on applied AI Research projects and innovations. Leading research projects, mentoring teams, and working collaboratively to drive advancements in technology.

Responsibilities

• Lead end-to-end applied research projects: define experiments, implement prototypes, run evaluations, and hand off reproducible artifacts to product/engineering teams.
• Own at least one deep technical domain (e.g., embedding model design & evaluation, context modeling for agentic systems, on-line or continuous adaptation / fine-tuning pipelines) while contributing across other areas.
• Translate research papers and state-of-the-art approaches into pragmatic, production-aware prototypes and AB-testable experiments.
• Build reliable evaluation pipelines and datasets (offline metrics + human evaluation) and document failures/lessons.
• Mentor and uplevel ML engineering practices across Mercari – techniques, model design, code quality, reproducible experiments, and sound evaluation.
• Collaborate closely with product managers, designers and platform engineers to scope and prioritize research that can move into product, as well as inspire and shape product vision.

Requirements

• Either one of the following:
• 5+ years building state-of-the-art ML systems in industry
• PhD and 1+ years industry experience
• Strong software engineering skills (Python, PyTorch or TF).
• Demonstrated depth and recent experience in at least one of our core focus areas:
• Semantic Understanding, e.g. multimodal-embeddings, representation learning, or latent variable modelling
• Contextual Intelligence, e.g. graph- or memory-augmented systems, retrieval-augmented generation, or agentic architectures that model user or item context
• Continuous Learning, e.g. preference learning (DPO/variants), reinforcement learning from user feedback (RLHF/RLAIF), or online/continual fine-tuning pipelines
• Strong track record of shipping prototypes or models end-to-end (not just research code).
• Ability to design experiments: dataset creation, metrics, human eval, and interpretable analysis.
• Excellent communication: explain technical tradeoffs to product and engineering audiences.
• Product & UX sense: willingness to frame user problems, success metrics, and UX trade‑offs with PM/Design.
• Comfortable working autonomously in a small, high-focus team that protects flow.
• Preferred Experience/Skills
• Evidence of autonomous research‑to‑product impact (open‑source libraries, internal platforms, or papers with code).
• Previous product or e-commerce experience (valuable but not required).
• Language
• English: Independent (CEFR - B2) Required
• Japanese: Independent (CEFR - B2) Optional
• For details about CEFR, see here.