Machine Learning Engineer（E2E Autonomous Driving Model Development）

必須

• 3+ years of model development experience using deep learning frameworks such as PyTorch
• Japanese language proficiency (JLPT N2 or equivalent)

• ◾️Plus at least one of the following:

  • Experience with Computer Vision (images, video) using deep learning
  • Experience developing ML models in a web or production setting using various approaches including deep learning
  • Experience with ML model development or quantization/optimization in autonomous driving, ADAS, or robotics

歓迎

• Award or top placement in ML competitions such as Kaggle
• Contribution experience to open-source ML libraries
• Knowledge of sensor fusion and sensor calibration
• Experience with unit testing and test-driven development (TDD)

求める人物像

• Driven to build a world-class company
• Self-starter who takes initiative on everything
• Humble, with genuine empathy for others
• Flexible and excited by rapid organizational and business growth
• Growth-oriented mindset
• Resilient — able to find joy even in tough challenges

使用ツール

Description

• This role is open to ML engineers with expertise in one or more of the following domains: Autonomous Driving, Computer Vision, or Machine Learning.
• At Turing, we are developing an End-to-End autonomous driving ML model — a single machine learning model that takes input from vehicle-mounted cameras and directly outputs vehicle control commands.
• Autonomous driving model development is a true multi-disciplinary challenge, spanning far beyond machine learning alone. There are many areas to contribute: data collection, dataset creation (data quality improvement, calibration, coordinate transforms), and model training (architecture design, training efficiency improvements), among others.
• We are looking for engineers with a background in autonomous driving — as well as engineers with outstanding expertise from software, robotics, or other industries. Let's tackle one of humanity's grand challenges together.

What you will work on

• We work on a wide range of problems — not just model architecture improvements, but also data quality and quantity challenges. The examples below are just a subset; if any of them resonate with your experience, we encourage you to apply.

Example：

• Implementation of End-to-End autonomous driving models
• Planning and strategy for data collection
• Dataset creation and improvement
• Auto-labeling model implementation and improvement
• Camera and sensor calibration
• Implementation of model training algorithms
• Optimization and speed-up of model training code
• On-vehicle model evaluation and experiment management
• Research, reproduction, and implementation of state-of-the-art papers

Our development approach

• We pursue both data-centric and model-centric approaches in parallel. Challenges arise from many angles — data quality issues with various root causes, architecture and backbone exploration — giving the team a wide solution space to work in. We also run large-scale training jobs on GPU clusters, so optimizing training code for speed is an active area of focus.
• E2E autonomous driving is still an open problem. The model you build could become the industry standard for the next generation of self-driving systems.

Test your model in the real world

• Our development cycle: Build dataset & model → Drive test → Analyze experiment logs → Manage model. You will iterate on your models by experiencing them firsthand in a real vehicle — not just on paper. Use feedback from the physical world to drive your development forward.

Tech stack

• Language：Python
• Libraries：PyTorch、OpenCV、MMDetection、ONNX、TensorRT
• Middleware：Slurm
• Cloud：AWS、（Databricks (DWH)）
• Platform：Jetson、Linux