業務内容

• Lead design and development of perception foundation models for autonomous vehicles, unifying diverse sensor data for scalable 3D scene understanding.
• Deploy scalable, efficient ML models on the vehicle platform with emphasis on safety standards and cost effectiveness.
• Build end-to-end ML pipelines: data strategy, preprocessing, training, ablations, evaluation, deployment, and inference optimization.
• Champion the scientific method and push for state-of-the-art deep learning solutions in a fast, agile environment.
• Collaborate closely with Perception, Motion Planning, Simulation, Infrastructure, and Tooling teams to deliver unified solutions.
• Work in a hybrid setup (Nihonbashi, Tokyo; Palo Alto; Ann Arbor) with on-site presence three days per week.

技術スタック

必須スキル

• MS or PhD in Machine Learning, Computer Vision, Robotics, or related quantitative field (or equivalent industry experience)
• 3+ years of Python, experience with major DL frameworks, and software engineering best practices
• 3+ years of DL approaches (supervised/unsupervised, transfer learning, multi-task, and/or deep reinforcement learning)
• 3+ years of ML workflow experience (data sampling/curation, preprocessing, training, ablations, evaluation, deployment, inference optimization)
• Strong understanding of runtime complexity, distributed/cloud ML infrastructure, data pipeline architecture, and resource-aware optimization
• Proficiency in C++ for integration with autonomous vehicle platform
• Strong leadership and communication skills

歓迎スキル（該当する場合）

• Published research at top-tier conferences (NeurIPS, CVPR) or equivalent
• Proven deployment of ML models at scale in self-driving or related fields
• Experience with production-level coding and deployment onto embedded platforms, optimizing for latency and hardware constraints
• Familiarity with large-scale foundation models, multimodal architectures, self-supervised learning
• Experience in perception/prediction/mapping/localization/planning/simulation

キャリア成長観点

• 直接影響を与える perception foundation modelsの設計・実装に携わり、実車での大規模データ活用とリアルタイム推論のスキルを深められる。
• データ戦略からトレーニング・評価・デプロイまでのエンドツーエンドなMLライフサイクルを横断的に経験できる。
• 安全性基準とハードウェア制約を考慮した実装・最適化を通じて、実装力とシステム設計力を同時に鍛えられる。
• 複数部門（Perception、Motion Planning、Simulation、Infrastructure、Tooling）との協働による技術戦略の形成とリーダーシップ機会がある。
• グローバル拠点でのハイブリッド勤務環境を活かし、ハードウェア/レイテンシ最適化やクラウド/エッジデプロイの経験を積み、研究成果の実用化可能性を高められる。